The Smartest Books We Know - Trailblazer CoachingΒ Β· Page 2 of 21 Contents Booms and Busts ..... 6
The Booms and Busts of Beta Arbitrage*personal.lse.ac.uk/loud/cobar.pdfThe Booms and Busts of Beta...
Transcript of The Booms and Busts of Beta Arbitrage*personal.lse.ac.uk/loud/cobar.pdfThe Booms and Busts of Beta...
The Booms and Busts of Beta Arbitrage*
Shiyang Huang University of Hong Kong
Email: [email protected]
Dong Lou London School of Economics and CEPR
Email: [email protected]
Christopher Polk London School of Economics and CEPR
Email: [email protected]
This Draft: March 2018
* We would like to thank Nicholas Barberis, Sylvain Champonnais, Andrea Frazzini, Pengjie Gao,
Emmanuel Jurczenko, Ralph Koijen, Toby Moskowitz, Lasse Pedersen, Steven Riddiough, Emil
Siriwardane, Dimitri Vayanos, Michela Verardo, Tuomo Vuolteenaho, Yu Yuan, and seminar participants
at Arrowstreet Capital, Citigroup Quant Research Conference, London Quant Group, London Business
School, London School of Economics, Norwegian School of Economics, Renmin University, UBS Quant
Conference, University of Cambridge, University of California in Los Angeles, University of Rotterdam,
University of Warwick, 2014 China International Conference in Finance, 2014 Imperial College Hedge Fund
Conference, 2014 Quantitative Management Initiative Conference, the 2015 American Finance Association
Conference, the 2015 Financial Intermediation Research Society Conference, the 2015 Northern Finance
Conference, the 2015 European Finance Association Conference, and the 2016 Finance Down Under
Conference for helpful comments and discussions. We are grateful for funding from the Europlace Institute
of Finance, the Paul Woolley Centre at the London School of Economics, and the QUANTVALLEY/FdR:
Quantitative Management Initiative.
The Booms and Busts of Beta Arbitrage
Abstract
Low-beta stocks deliver high average returns and low risk relative to high-beta
stocks, an opportunity for professional investors to βarbitrageβ away. We argue
that beta-arbitrage activity instead generates booms and busts in the strategyβs abnormal trading profits. In times of low arbitrage activity, the beta-arbitrage
strategy exhibits delayed correction, taking up to three years for abnormal returns to be realized. In stark contrast, when activity is high, prices overshoot as short-
run abnormal returns are much larger and then revert in the long run. We
document a novel positive-feedback channel operating through firm-level leverage that facilitates these boom and bust cycles. These cyclical patterns also show up
in hedge fund exposures to beta arbitrage, particularly exposures of smaller and thus more nimble funds.
1
I. Introduction
The trade-off of risk and return is a key concept in modern finance. The simplest and
most intuitive measure of risk is market beta β the slope in the regression of a securityβs
return on the market return. In the Capital Asset Pricing Model (CAPM) of Sharpe (1964)
and Lintner (1965), market beta is the only risk needed to explain expected returns. More
specifically, the CAPM predicts that the relation between expected return and beta β the
security market line β has an intercept equal to the risk-free rate and a slope equal to the
equity premium.
However, empirical evidence indicates that the security market line is too flat on
average (Black 1972, Frazzini and Pedersen, 2014) and especially so during times of high
expected inflation (Cohen, Polk, and Vuolteenaho 2005), disagreement (Hong and Sraer
2016), and market sentiment (Antoniou, Doukas, and Subrahmanyam 2015). These
patterns are not explained by other well-known asset pricing anomalies such as size, value,
and price momentum.
We study the response of arbitrageurs to this failure of the Sharpe-Lintner CAPM
in order to identify booms and busts of beta arbitrage.1 In particular, we exploit the novel
measure of arbitrage activity introduced by Lou and Polk (2014). They argue that
traditional measures of such activity are flawed, poorly measuring a portion of the inputs
to the arbitrage process, for a subset of arbitrageurs. Lou and Polkβs innovation is to
measure the outcome of the arbitrage process, namely, the correlated price impacts that
can result in excess return comovement in the spirit of Barberis and Shleifer (2003).2
We first confirm that our measure of the excess return comovement, relative to a
benchmark asset pricing model, of beta-arbitrage stocks (labelled πΆππ΅π΄π ) is correlated
1 Schwert (2003) summarizes the performance of many trading strategies and motivated by his results,
argues that practitioners who trade on anomalies after their publication can cause the effects to disappear.
McLean and Pontiff (2016) systematically study the out-of-sample and post-publication performance for 97
variables shown to predict stock returns.
2 See, for example, Barberis, Shleifer and Wurgler (2005), Greenwood and Thesmar (2011), Lou (2012) and
Anton and Polk (2014).
2
with existing measures of arbitrage activity. In particular, we find that time variation in
the level of institutional holdings in low-beta stocks (i.e., stocks in the long leg of the beta
strategy), the assets under management of long-short equity hedge funds, aggregate
liquidity, and the past performance of a typical beta-arbitrage strategy together forecast
roughly 35% of the time-series variation in πΆππ΅π΄π . These findings suggest that not only
is our measure consistent with existing proxies for arbitrage activity but also that no one
single existing proxy is sufficient for capturing time-series variation in arbitrage activity.
Indeed, one could argue that perhaps much of the unexplained variation in πΆππ΅π΄π
represents variation in arbitrage activity missed by existing measures.
After validating our measure in this way, we then forecast the cumulative abnormal
returns to beta arbitrage. We first find that when arbitrage activity is relatively high (as
identified by the 20% of the sample period with the highest values of πΆππ΅π΄π ), abnormal
returns to beta-arbitrage strategies occur relatively quickly, within the first six months of
the trade. In contrast, when arbitrage activity is relatively low (as identified by the 20%
of the sample period with the lowest values of πΆππ΅π΄π ), any four-factor abnormal returns
to beta-arbitrage strategies take much longer to materialize, appearing three years after
putting on the trade.
These effects are both economically and statistically significant. When beta-
arbitrage activity is low, the abnormal four-factor returns on beta arbitrage are
statistically insignificant from zero in the two years after portfolio formation. For the
patient arbitrageur, in year 3, the strategy earns abnormal four-factor returns of 0.59%
per month with a t-statistic of 2.22. In stark contrast, for those periods when arbitrage
activity is high, the abnormal four-factor returns to beta arbitrage average 1.08% per
month with a t-statistic of 2.63 in the six months after the trade.
We then show that the stronger performance of beta-arbitrage activities during
periods of high beta-arbitrage activity can be linked to subsequent reversal of those profits.
In particular, the year 3 abnormal four-factor returns are -0.93% with an associated t-
3
statistic of -2.66. As a consequence, the long-run reversal of beta-arbitrage returns varies
predictably through time in a striking fashion. The post-formation, year-3 spread in
abnormal four-factor returns across periods of low arbitrage activity, when abnormal
returns are predictably positive, and periods of high arbitrage activity, when abnormal
returns are predictably negative, is -1.52%/month (t-statistic = -3.33) or nearly 20%
cumulative in that year.
Our results reveal interesting patterns in the relation between arbitrage returns
and the arbitrage crowd. When beta-arbitrage activity is low, the returns to beta-arbitrage
strategies exhibit significant delayed correction. In contrast, when beta-arbitrage activity
is high, the returns to beta-arbitrage activities reflect strong over-correction due to
crowded arbitrage trading. These results are consistent with time-varying arbitrage
activity generating booms and busts in beta arbitrage.
We argue that these results are intuitive, as it is difficult to know how much
arbitrage activity is pursuing beta arbitrage, and, in particular, the strategy is susceptible
to positive-feedback trading. Specifically, successful bets on (against) low-beta (high-beta)
stocks result in prices for those securities rising (falling). If the underlying firms are
leveraged, this change in price will, all else equal, result in the securityβs beta falling
(increasing) further. Thus, not only do arbitrageurs not know when to stop trading the
low-beta strategy, their (collective) trades strengthen the signal. Consequently, beta
arbitrageurs may increase their bets precisely when trading is more crowded.3
Consistent with our novel positive-feedback mechanism, we show that the cross-
sectional spread in betas increases when beta-arbitrage activity is high and particularly
so when beta-arbitrage stocks are relatively more levered. As a consequence, stocks remain
in the extreme beta portfolios for a longer period of time. Our novel positive feedback
channel also has implications for cross-sectional heterogeneity in abnormal returns; we
3 Note that crowded trading may or may not be profitable, depending on how long the arbitrageur holds
the position and how long it takes for any subsequent correction to occur.
4
find that our boom and bust beta-arbitrage cycles are particularly strong among high-
leverage stocks.
A variety of robustness tests confirm our main findings. In particular, we show
that controlling for other factors when either measuring πΆππ΅π΄π or when predicting beta-
arbitrage returns does not alter our primary conclusions a) that the excess comovement
of beta-arbitrage stocks forecasts time-varying reversal to beta-arbitrage bets and b) that
the beta spread varies with πΆππ΅π΄π .
Our findings can also be seen by estimating time variation in the short-run (months
1-6) and long-run (year 3) security market lines, conditioning on πΆππ΅π΄π . Thus, the
patterns we find are not just due to extreme-beta stocks, but reflect dynamic movements
throughout the entire cross section. In particular, we find that during periods of high beta-
arbitrage activity, the short-term security market line strongly slopes downward,
indicating strong profits to the low-beta strategy, consistent with arbitrageurs expediting
the correction of market misevaluation. However, this correction is excessive, as the long-
run security market line dramatically slopes upwards. In contrast, during periods of low
beta-arbitrage activity, the short-term security market line is weakly upward sloping.
During these low-arbitrage periods, we do not find any downward slope to the security
market line until the long-run.
A particularly compelling robustness test involves separating πΆππ΅π΄π into excess
comovement among low-beta stocks occurring when these stocks have relatively high
returns (i.e., capital flowing into low beta stocks and pushing up the prices) vs. excess
comovement occurring when low-beta stocks have relatively low returnsβi.e., upside
versus downside comovement. Under our interpretation of the key findings, it is the former
that should track time-series variation in expected beta-arbitrage returns, as that
particular direction of comovement is consistent with trading aiming to correct the beta
anomaly. Though estimates of upside πΆππ΅π΄π are naturally much noisier, our evidence
confirms the intuition above: our main results are primarily driven by upside πΆππ΅π΄π .
5
Finally, Shleifer and Vishny (1997) link the extent of arbitrage activity to limits
to arbitrage. Based on their logic, trading strategies that bet on firms that are cheaper to
arbitrage (e.g., larger stocks, more liquid stocks, or stocks with lower idiosyncratic risk)
should have more arbitrage activity. This idea of limits to arbitrage motivates tests
examining cross-sectional heterogeneity in our findings. We show that our results
primarily occur in those stocks that provide the least limits to arbitrage: large stocks,
liquid stocks, and stocks with low idiosyncratic volatility. This cross-sectional
heterogeneity in the effect is again consistent with the interpretation that arbitrage
activity causes much of the time-varying patterns we document.
With these patterns in hand, we take a closer look at how our measure reflects
investment activity in this strategy in the cross-section of mutual funds and hedge funds.
Specifically, we show that the typical long-short equity hedge fund increases its beta-
arbitrage exposure when πΆππ΅π΄π is relatively high (that is, when short-term beta arbitrage
returns are also higher). However, the ability of hedge funds to time the strong
overreaction that occurs when πΆππ΅π΄π is high declines with assets under management.
During booms in beta-arbitrage, small hedge funds have positive exposures to a low-beta
factor that are more than twice as big as their large fund counterparts. In stark contrast,
mutual funds have an insignificant negative exposure to low-beta strategies (i.e., they
overweight high-beta stocks), which does not vary with πΆππ΅π΄π .
The organization of our paper is as follows. Section II summarizes the related
literature. Section III describes the data and empirical methodology. We detail our
empirical findings regarding beta-arbitrage activity and predictable patterns in returns in
section IV, and present key tests of our economic mechanism in Section V. Section VI
concludes.
II. Related Literature
Our results shed new light on the risk-return trade-off, a cornerstone of modern asset
pricing research. This trade-off was first established in the Sharpe-Lintner CAPM, which
6
argues that the market portfolio is mean-variance efficient. Consequently, a stockβs
expected return is a linear function of its market beta, with a slope equal to the equity
premium and an intercept equal to the risk-free rate.
However, mounting empirical evidence is inconsistent with the CAPM. Black,
Jensen, and Scholes (1972) were the first to show carefully that the security market line
is too flat on average. Put differently, the risk-adjusted returns of high beta stocks are
too low relative to those of low-beta stocks. This finding was subsequently confirmed in
an influential study by Fama and French (1992). Blitz and van Vliet (2007) and Baker,
Bradley, and Taliaferro (2014), Frazzini and Pedersen (2014), and Blitz, Pang, and van
Vliet (2013) document that the low-beta anomaly is also present in both non-US developed
markets as well as emerging markets.
Of course, the flat security market line is not the only failing of the CAPM (see
Fama and French 1992, 1993, and 1996). Nevertheless, since this particular issue is so
striking, a variety of explanations have been offered to explain the low-beta phenomenon.
Black (1972) and more recently Frazzini and Pedersen (2014) argue that leverage-
constrained investors, such as mutual funds, tend to deviate from the capital market line
and invest in high beta stocks to pursue higher expected returns, thus causing these stocks
to be overpriced relative to the CAPM benchmark.4,5,6
Cohen, Polk, and Vuolteenaho (2005) derive the cross-sectional implications of the
CAPM in conjunction with the money illusion story of Modigliani and Cohn (1979). They
show that money illusion implies that, when inflation is low or negative, the compensation
for one unit of beta among stocks is larger (and the security market line steeper) than the
rationally expected equity premium. Conversely, when inflation is high, the compensation
4 JylhΓ€ (2017) provides evidence in support of this interpretation using Federal Reserve changes in initial
margin requirements.
5 See also Karceski (2002), Baker, Bradley, and Wurgler (2011) and Buffa, Vayanos, and Woolley (2014)
for related explanations based on benchmarking of institutional investors.
6 Koijen and Yogo (2017) provide a general framework for modeling the role of institutions in asset markets.
7
for one unit of beta among stocks is lower (and the security market line shallower) than
what the overall pricing of stocks relative to bills would suggest. Cohen, Polk, and
Vuolteenaho provide empirical evidence in support of their theory.
Hong and Sraer (2016) provide an alternative explanation based on Millerβs (1977)
insights. In particular, they argue that investors disagree about the value of the market
portfolio. This disagreement, coupled with short sales constraints, can lead to
overvaluation, and particularly so for high-beta stocks, as these stocks allow optimistic
investors to tilt towards the market. Further, Kumar (2009) and Bali, Cakici, and
Whitelaw (2011) show that high risk stocks can indeed underperform low risk stocks, if
some investors prefer volatile, skewed returns, in the spirit of the cumulative prospect
theory as modeled by Barberis and Huang (2008). Building on arguments in Stambaugh,
Yu, and Yuan (2015), Liu, Stambaugh, and Yuan (2017) attribute the beta anomaly to
the positive correlation between market beta and idiosyncratic volatilities.7
A natural question is why sophisticated investors, who can lever up and sell short
securities at relatively low costs, do not fully take advantage of this anomaly and thus
restore the theoretical relation between risk and returns. Our paper is aimed at addressing
this exact question. Our premise is that professional investors indeed take advantage of
this low-beta return pattern, often in dedicated strategies that buy low-beta stocks and
sell high-beta stocks. However, the amount of capital that is dedicated to this low-beta
strategy is both time varying and unpredictable from arbitrageursβ perspectives, thus
resulting in periods where the security market line remains too flatβi.e., too little
arbitrage capital, as well as periods where the security market line becomes overly steepβ
i.e., too much arbitrage capital.
Not all arbitrage strategies have these issues. Indeed, some strategies have a natural
anchor that is relatively easily observed (Stein 2009). For example, it is straightforward
7 In addition, Campbell, Giglio, Polk, and Turley (2018) document that high-beta stocks hedge time-
variation in the aggregate marketβs return volatility, offering a potential neoclassical explanation for the
low-beta anomaly.
8
to observe the extent to which an ADR is trading at a price premium (discount) relative
to its local share. This ADR premium/discount is a clear signal to an arbitrageur of an
opportunity and, in fact, arbitrage activity keeps any price differential small with
deviations disappearing within minutes.8 Importantly, if an unexpectedly large number of
ADR arbitrageurs pursue a particular trade, the price differential narrows. An individual
ADR arbitrageur can then adjust his or her demand accordingly.
There is, however, no easy anchor for beta arbitrage.9 Further, we argue that the
difficulty in identifying the amount of beta-arbitrage capital is exacerbated by a novel,
indirect positive-feedback channel.10 Namely, beta-arbitrage trading can lead to the cross-
sectional beta spread increasing when firms are levered. As a consequence, stocks in the
extreme beta deciles are more likely to remain in these extreme groups, with more extreme
beta values, when arbitrage trading becomes excessive. Given that beta arbitrageurs rely
on realized beta as their trading signal, this beta expansion resulting from leverage
effectively causes a positive feedback loop in the beta-arbitrage strategy.
III. Data and Methodology
The main dataset used in this study is the stock return data from the Center for Research
in Security Prices (CRSP). Following prior studies on the beta-arbitrage strategy, we
include in our study all common stocks on NYSE, Amex, and NASDAQ. We then
8 RΓΆsch (2014) studies various properties of ADR arbitrage. For his sample of 72 ADR home stock pairs,
the average time it takes until a ADR/home stock price deviation disappears is 252 seconds. For an
institutional overview of this strategy, see J.P. Morgan (2014).
9 Polk, Thompson, and Vuolteenaho (2006) use the Sharpe-Lintner CAPM to relate the cross-sectional beta
premium to the equity premium. They show how the divergence of the two types of equity-premium
measures implies a time-varying trading opportunity for beta arbitrage. Their methods are quite
sophisticated and produce signals about the time-varying attractiveness of beta-arbitrage that, though useful
in predicting beta-arbitrage returns, are still, of course, quite noisy.
10 The idea that positive-feedback strategies are prone to destabilizing behaviour goes back to at least
DeLong, Shleifer, Summers, and Waldmann (1990). In contrast, negative-feedback strategies like ADR
arbitrage or value investing are less susceptible to destabilizing behaviour by arbitrageurs, as the price
mechanism mediates any potential congestion. See Stein (2009) for a discussion of these issues.
9
augment the stock return data with institutional ownership in individual stocks provided
by Thompson Financial. We further obtain information on assets under management of
long-short equity hedge funds from Lipperβs Trading Advisor Selection System (TASS).
Since the assets managed by hedge funds grow substantially in our sample period, we
detrend this variable. In addition, we use fund-level data on hedge fund returns and AUM.
We also construct, as controls, a list of variables that have been shown to predict
future beta-arbitrage strategy returns. Specifically, a) following Cohen, Polk, and
Vuolteenaho (2005), we construct a proxy for expected inflation using an exponentially
weighted moving average (with a half-life of 36 months) of past log growth rates of the
producer-price index; b) we also include in our study the sentiment index proposed by
Baker and Wurgler (2006, 2007); c) following Hong and Sraer (2016), we construct an
aggregate disagreement proxy as the beta-weighted standard deviation of analystsβ long-
term growth rate forecasts; d) finally, following Frazzini and Pedersen (2014), we use the
Ted spreadβthe difference between the LIBOR rate and the US Treasury bill rateβas a
measure of financial intermediariesβ funding constraints.
We begin our analysis in January 1970 (i.e., our first measure of beta arbitrage
crowdedness is computed as of December 1969), as that was when the low-beta anomaly
was first recognized by academics.11 At the end of each month, we sort all stocks into
deciles (in some cases vigintiles) based on their pre-ranking market betas. Following prior
literature, we calculate pre-ranking betas using daily returns in the past twelve months
(with at least 200 daily observations). Our results are similar if we use monthly returns,
or different pre-ranking periods. To account for illiquidity and non-synchronous trading,
we include on the right hand side of the regression equation five lags of the excess market
return, in addition to the contemporaneous excess market return. The pre-ranking beta is
simply the sum of the six coefficients from the OLS regression.
11 Though eventually published in 1972, Black, Jensen, and Scholes had been presented as early as August
of 1969. Mehrlingβs (2005) biography of Fischer Black details the early history of the low-beta anomaly.
10
We then compute pairwise partial correlations using 52 (non-missing) weekly
returns for all stocks in each decile in the portfolio ranking period. We control for the
Fama-French three factors when computing these partial correlations to purge out any
comovement in stocks induced by known risk factors. We measure the excess comovement
of stocks involved in beta arbitrage (πΆππ΅π΄π ) as the average pairwise partial correlation
in the lowest market beta decile. We focus on the low-beta decile as these stocks tend to
be larger, more liquid, and have lower idiosyncratic volatility compared to the highest-
beta decile; thus, our measurement of excess comovement will be less susceptible to issues
related to asynchronous trading and measurement noise. 12 We operationalize this
calculation by computing the average correlation of the three-factor residual of every stock
in the lowest beta decile with the rest of the stocks in the same decile:
πΆππ΅π΄π =1
πβ ππππ‘ππππΆπππ(πππ‘πππ
πΏ , πππ‘ππβππΏ |πππ‘ππ, π ππ, βππ)
π
π=1
,
where πππ‘ππππΏ is the weekly return of stock π in the (L)owest beta decile, πππ‘ππβπ
πΏ is the
weekly return of the equal-weight lowest beta decile excluding stock π , and π is the
number of stocks in the lowest beta decile. We have also measured πΆππ΅π΄π using returns
that are orthogonalized not only to the Fama-French factors but also to each stockβs
industry return or to other empirical priced factors, and our conclusions continue to hold.
We present these and many other robustness tests in Table IV.
In the following period, we then form a zero-cost portfolio that goes long the value-
weight portfolio of stocks in the lowest market beta decile and short the value-weight
portfolio of stocks in the highest market beta decile. We track the cumulative abnormal
returns of this zero-cost long-short portfolio in months 1 through 36 after portfolio
formation. To summarize the timing of our empirical exercise, year 0 is our portfolio
formation year (during which we also measure πΆππ΅π΄π ), year 1 is the holding year, and
12 Our results are robust to measuring πΆππ΅π΄π using a pooled sample of high- and low-beta stocks (putting
a negative sign in front of the returns of high-beta stocks), as well as the (minus) cross-correlation between
high- and low-beta deciles.
11
years 2 and 3 are our post-holding period, to detect any (conditional) long-run reversal to
the beta-arbitrage strategy.
IV. Main Results
We first document simple characteristics of our arbitrage activity measure. Table I Panel
A indicates that there is significant excess correlation among low-beta stocks on average
and that this pairwise correlation varies substantially through time; specifically, the mean
of πΆππ΅π΄π is 0.11 varying from a low of 0.04 to a high of 0.20.
Panel B of Table I examines πΆππ΅π΄π βs correlation with existing measures linked to
time variation in the expected abnormal returns to beta-arbitrage strategies. We find that
πΆππ΅π΄π is high when disagreement is high, with a correlation of 0.34. πΆππ΅π΄π is also
positively correlated with the Ted spread, consistent with a time-varying version of Black
(1972), though the Ted spread does not forecast (or in some cases forecasts in the wrong
direction) time variation in expected abnormal returns to beta-arbitrage strategies
(Frazzini and Pederson 2014). πΆππ΅π΄π is negatively correlated with the expected inflation
measure of Cohen, Polk, and Vuolteenaho. However, in results not shown, the correlation
between expected inflation and πΆππ΅π΄π becomes positive for the subsample from 1990-
2010, consistent with arbitrage activity eventually taking advantage of this particular
source of time-variation in beta-arbitrage profits. There is little to no correlation between
πΆππ΅π΄π and sentiment.
Figure 1 plots πΆππ΅π΄π as of the end of each December. Note that we do not
necessarily expect a trend in this measure. Though there is clearly more capital invested
in beta-arbitrage strategies, in general, markets are also deeper and more liquid.
Nevertheless, after an initial spike in December 1971, πΆππ΅π΄π trends slightly upward for
the rest of the sample. However, there are clear cycles around this trend. These cycles
tend to peak before broad market declines. Also, note that πΆππ΅π΄π is essentially
12
uncorrelated with market volatility. A regression of πΆππ΅π΄π on contemporaneous realized
market volatility produces a loading of -0.23 with a t-statistic of -0.79.
Consistent with our measure tracking arbitrage activity, Appendix Table A1 shows
that πΆππ΅π΄π is persistent in event time. Specifically, the correlation between πΆππ΅π΄π
measured in year 0 and year 1 for the same set of stocks is 0.15. In fact, year-0 πΆππ΅π΄π
remains highly correlated with subsequent values of πΆππ΅π΄π for the same stocks all the
way out to year 3. The average value of πΆππ΅π΄π remains high as well. Recall that in year
0, the average excess correlation is 0.11. We find that in years 1, 2, and 3, the average
excess correlation of these same stocks remains around 0.07.13
IV.A. Determinants of πΆππ΅π΄π
To confirm that our measure of beta-arbitrage is sensible, we estimate regressions
forecasting πΆππ΅π΄π with four variables that are often used to proxy for arbitrage activity.
The first variable we use is the aggregate institutional ownership (Inst Own) of the low-
beta decileβi.e., stocks in the long leg of the beta strategyβbased on 13F filings. We
include institutional ownership as these investors are typically considered smart money,
at least relative to individuals, and we focus on their holdings in the low-beta decile as
we do not observe their short positions in the high-beta decile. We also include the assets
under management (AUM) of long-short equity hedge funds, the prototypical arbitrageur.
Finally, we include a measure of the past profitability of beta-arbitrage strategies, the
realized four-factor alpha of Frazzini and Pedersenβs BAB factor. Intuitively, more
arbitrageurs should be trading the low-beta strategy after the strategy has performed well
in recent past.
All else equal, we expect πΆππ΅π΄π to be lower if markets are more liquid. However,
as arbitrage activity is endogenous, times when markets are more liquid may also be times
13 πΆππ΅π΄π is essentially uncorrelated with a similar measure of excess comovement based on the fifth and
sixth beta deciles.
13
when arbitrageurs are more active. Indeed, Cao, Chen, Liang, and Lo (2013) show that
hedge funds increase their activity in response to increases in aggregate liquidity.
Following Cao, Chen, Liang, and Lo, we further include past market liquidity as proxied
by the Pastor and Stambaugh (2003) liquidity factor (PS liquidity) in our regressions to
measure which channel dominates.
All regressions in Table II include a trend to ensure that our results are not
spurious. We also report a second specification that includes variables that arguably
should forecast beta-arbitrage returns: the inflation, sentiment, and disagreement indices
as well as the Ted spread. We measure these variables contemporaneously with πΆππ΅π΄π
as we will be running horse races against these variables in our subsequent analysis.
Regression (1) in Table II documents that Inst Own, AUM, and PS liquidity forecast
πΆππ΅π΄π , with an R2 of approximately 35%.14 Regression (2) shows that two of the extant
predictors of beta-arbitrage returns help explain πΆππ΅π΄π . The Ted spread, adds some
incremental explanatory power, with the sign of the coefficient consistent with
arbitrageurs taking advantage of potential time-variation in beta-arbitrage returns linked
to this channel. Indeed, as we show later, the Ted spread does a poor job forecasting beta-
arbitrage returns in practice, perhaps because arbitrageurs have compensated
appropriately for this potential departure from Sharpe-Lintner pricing. The disagreement
measure also helps explain variation in πΆππ΅π΄π . Finally, in this full specification, past
profitability of a prototypical beta-arbitrage strategy strongly forecasts relatively high
arbitrage activity going forward. It seems reasonable that strong past performance of an
investment strategy may result in the strategy becoming more popular.
Overall, these findings make us comfortable in our interpretation that πΆππ΅π΄π is
related to arbitrage activity and distinct from existing measures of opportunities in beta
14 We choose to forecast πΆππ΅π΄π in a predictive regression rather than explain πΆππ΅π΄π in a contemporaneous
regression simply to reduce the chance of a spurious fit. However, our results are robust to estimating
contemporaneous versions of these regressions.
14
arbitrage. As a consequence, we turn to the main analysis of the paper, the short- and
long-run performance of beta-arbitrage returns, conditional on πΆππ΅π΄π .
IV.B. Forecasting Beta-Arbitrage Returns
Table III forecasts the abnormal returns on the standard beta-arbitrage strategy as a
function of investment horizon, conditional on πΆππ΅π΄π . Panel A examines Fama and
French (1993) three-factor-adjusted returns while Panel B studies abnormal returns
relative to the four-factor model of Carhart (1997).15 In each panel, we measure the
average abnormal returns in the first six months subsequent to the beta-arbitrage trade,
months seven through 12, and then those occurring in years two and three. We also report
the average abnormal returns across years two and three combined. These returns are
measured as a function of the value of πΆππ΅π΄π as of the end of the beta formation period.
In particular, we split the sample into five πΆππ΅π΄π quintiles.
We focus on the four-factor estimates. Pursuing beta arbitrage when arbitrage
activity is low takes patience. Abnormal four-factor returns are statistically insignificant
in the first year for the bottom four πΆππ΅π΄π groups.16 Abnormal returns only become
statistically significant for the lowest πΆππ΅π΄π group in the third year. In the lowest πΆππ΅π΄π
period, the four-factor alpha is 0.59%/month with an associated t-statistic of 2.22.17
However, as beta-arbitrage activity increases, the abnormal returns arrive sooner
and stronger. For the highest πΆππ΅π΄π group, the abnormal four-factor returns average
15 We compute the four-factor alpha for each event month after portfolio formation by taking the average
across all the calendar months in our sample.
16 πΆππ΅π΄π has a non-linear, U-shaped relation with beta arbitrage returns in the following year. As can be
seen in Appendix Tables A2 and A3, once we include both πΆππ΅π΄π and πΆππ΅π΄π 2 in the regression to forecast
future beta arbitrage returns, both coefficients are statistically and economically significant, but have
opposite signs. In other words, beta arbitrage returns are relatively high following periods of both extreme
high arbitrage activity (when arbitrageurs collectively bet against mispricing) and extreme low arbitrage
activity (when great opportunities are left unexploited).
17 We have also separately examined the long and short legs of beta arbitrage (i.e., low-beta vs. high-beta
stocks). Around 40% of our return effect comes from the long leg, and the remaining 60% from the short
leg.
15
1.08%/month in the six months immediately subsequent to the beta-arbitrage trade. This
finding is statistically significant with a t-statistic of 2.63, though the difference between
abnormal returns in high and low πΆππ΅π΄π periods, despite being economically large at
0.75%/month, is statistically insignificant. (In a regression context as in table V, where
we have more power to isolate the effect of crowded trading, πΆππ΅π΄π positively forecasts
beta arbitrage returns in the first six months.)
The key finding of our paper is that these quicker and stronger beta-arbitrage
returns can be linked to subsequent reversal in the long run. Specifically, in year three,
the abnormal four-factor return to beta arbitrage when πΆππ΅π΄π is high is -0.93%/month,
with a t-statistic of -2.66. These abnormal returns are dramatically different from their
corresponding values when πΆππ΅π΄π is low; the difference in year 3 abnormal four-factor
returns is -1.52%/month (t-statistic = -3.33).18
Since splitting the long run at year 2 is arbitrary, Table III also reports the results
combining years 2 and 3 together. The patterns remain, and these estimates are
monotonically decreasing in πΆππ΅π΄π .
Figure 2 summarizes these patterns by plotting the cumulative abnormal four-
factor returns to beta arbitrage during periods of high and low πΆππ΅π΄π , accumulating
abnormal returns up to 60 months post portfolio formation. We include in the plot the
cumulative four-factor returns during median πΆππ΅π΄π periods as well. This figure clearly
shows that there is a significant delay in abnormal trading profits to beta arbitrage when
beta-arbitrage activity is low. However, when beta-arbitrage activity is high, beta
arbitrage results in price overshooting, as evidenced by the initial price run-up and
subsequent reversal that we document. We argue that trading of the low-beta anomaly is
initially stabilizing, then, as the trade becomes crowded, turns destabilizing, causing prices
to overshoot. The bottom panel of Figure 2 further shows that the run-up to the beta
18 We postpone the discussion of conditional abnormal returns to beta arbitrage (as shown in the last row
of Table II) to Section V.C.
16
arbitrage strategy during high πΆππ΅π΄π periods starts in the formation period, consistent
with the view that arbitrageurs may use shorter windows to calculate beta. If this
formation-period runup is included in the accumulation of abnormal return, we can
account for effectively all of the reversal.
IV.C. Robustness of Key Results
Table IV examines variations to our methodology to ensure that our finding of time-
varying reversal of beta-arbitrage profits is robust. For simplicity, we report the difference
in returns to the beta strategy between the high and low πΆππ΅π΄π groups in two versions
of the long run (year 3 and years 2-3). For reference, the first row of Table IV reports the
baseline results from Table III Panel B.
In row two, we conduct the same analysis for the sub-period before our sample
(1927-1969). Of course, this sample not only predates the discovery of the low-beta
anomaly but also is a period where there is much less arbitrage activity in general, at
least explicitly organized as such. Thus, this period could be thought of as a placebo test
of our story. Consistent with our paperβs explanation, we find no statistically significant
link between πΆππ΅π΄π and the reversal of beta-arbitrage returns.
Our remaining subsample analysis excludes potential outlier years. We find that
our results remain robust if we exclude the tech bubble crash (2001) or the recent financial
crisis (2007-2009) from our sample.
In rows five through eight, we report the results from similar tests using extant
variables linked to potential time variation in beta-arbitrage profits. None of the four
variables are associated with time variation in long-run abnormal returns. Thus, our
πΆππ΅π΄π measure has not simply repackaged an effect linked to existing forecasting
variables.
In rows nine through 15, we consider alternative definitions of πΆππ΅π΄π . In the ninth
row, we control for UMD when computing πΆππ΅π΄π . In row 10, we separate HML into its
17
large cap and small cap components. In Row 11, we report results based on Fama and
French (2015a, 2015b) five-factor returns. In Row 12, we perform the entire analysis on
an industry-adjusted basis by sorting stocks into beta deciles within industries. Row 13
uses the correlation between the high-beta and low-beta deciles as a measure of arbitrage
activity, with lower values indicating more activity.19
Rows 14 and 15 split πΆππ΅π΄π into upside and downside components. Specifically,
we measure the following
πΆππ΅π΄π π =1
πβ ππππ‘ππππΆπππ(πππ‘πππ
πΏ, πππ‘ππβππΏ |πππ‘ππ, π ππ, βππ, πππ‘πππΏ > ππππππ(πππ‘πππΏ))
π
π=1
πΆππ΅π΄π π· =1
πβ ππππ‘ππππΆπππ(πππ‘πππ
πΏ, πππ‘ππβππΏ |πππ‘ππ, π ππ, βππ, πππ‘πππΏ < ππππππ(πππ‘πππΏ))
π
π=1
Separating πΆππ΅π΄π in this way allows us to distinguish between excess comovement tied
to strategies buying low-beta stocks (such as those followed by beta arbitrageurs) and
strategies selling low-beta stocks (such as leveraged-constrained investors modeled by
Black (1972)). Consistent with our interpretation, we find that only πΆππ΅π΄π π forecasts
time variation in the short- and long-run expected returns to beta arbitrage (whereas
πΆππ΅π΄π π· does not).
Rows 16-22 document that our results are robust to replacing πΆππ΅π΄π with
residual πΆππ΅π΄π . In particular, we orthogonalize πΆππ΅π΄π to measures of arbitrage activity
in momentum and value (Lou and Polk 2014), the average correlation in the market
(Pollet and Wilson 2010), the past volatility of beta-arbitrage returns, the volatility of
market returns over the twelve-month period corresponding to the measurement of
πΆππ΅π΄π , a trend, and lagged πΆππ΅π΄π (the year -1 average pairwise excess correlation of
low-beta stocks identified in year 0). Finally, in row 23, we measure abnormal returns
19 We also compute πΆππ΅π΄π based on a pooled sample of high- and low-beta stocks, putting a negative sign
in front of the returns of high-beta stocks. The results are similar. For example, after controlling for other
confounding factors, the difference in three-factor alpha of beta arbitrage in year three after portfolio
formation between high and low πΆππ΅π΄π periods is -0.96%/month (t-statistic = -2.24), and that same
difference in four-factor alpha is -0.77%/month (t-statistic = -1.79).
18
using a six-factor model that augments the Fama-French five-factor model with
momentum.
In all cases, πΆππ΅π΄π continues to predict time-variation in year 3 returns. The
estimates are always economically significant, with most point estimates larger than
1%/month. Statistical significance is always strong as well, with most t-statistic larger
than 3. Taken together, these results confirm that our measure of crowded beta arbitrage
robustly forecasts times of strong reversal to beta-arbitrage strategies.
In Table V, we report the results of regressions forecasting the abnormal four-factor
returns to beta-arbitrage spread bets, while controlling for other predictors of beta-
arbitrage returns. Unlike Table II, these regressions exploit not just the ordinal but also
the cardinal aspect of πΆππ΅π΄π . Moreover, these regressions not only confirm that our
findings are robust to existing measures of the profitability of beta arbitrage, they also
document the relative extent to which existing measures forecast abnormal returns to
beta-arbitrage strategies in the presence of πΆππ΅π΄π .
Regressions (1)-(3) in Table V forecast time-series variation in abnormal beta-
arbitrage returns in months 1-6. Regression (1) confirms that πΆππ΅π΄π strongly forecasts
beta-arbitrage four-factor alphas over the full sample. Regression (2) then includes
controls that are available over the entire sample. These include the inflation and
sentiment indices, market volatility, and a version of Cohen, Polk, and Vuolteenahoβs
(2003) value spread for the beta deciles in question. πΆππ΅π΄π continues to reliably describe
time-variation in abnormal four-factor returns on the low-beta-minus-high-beta strategy,
with only the sentiment index providing any additional explanatory power. Over the
shorter sample period where both aggregate disagreement and the Ted spread are
available, πΆππ΅π΄π retains its economic significance, but becomes marginally statistically
significant in forecasting time-variation in the abnormal returns to the beta-arbitrage
strategy.
19
Regressions (4)-(6) of Table V forecast the returns on beta-arbitrage strategies in
year 3. The message from these regressions concerning the main result of the paper is
clear; πΆππ΅π΄π strongly forecasts a time-varying reversal, while none of the extant variables
has any predictive power for long-run buy-and-hold beta-arbitrage returns.
IV.D. Predicting the Security Market Line
Our results can also be seen from time variation in the shape of the security market line
(SML) as a function of lagged πΆππ΅π΄π . Such an approach can help ensure that the time-
variation we document is not restricted to a small subset of extreme-beta stocks, but
instead is a robust feature of the cross-section. (We argue that beta arbitrage activity can
affect the entire cross-section of stocks rather than just the extreme deciles, because
arbitrageurs may bet against the low-beta anomalies by selecting portfolio weights that
are inversely proportional to the market beta.) At the end of each month, we sort all
stocks into 20 value-weighted portfolios by their pre-ranking betas.20 We track these 20
portfolio returns both in months 1-6 and months 25-36 after portfolio formation to
compute short-term and long-term post-ranking betas, and, in turn, to construct our
short-term and long-term security market lines.
For the months 1-6 portfolio returns, we then compute the post-ranking betas by
regressing each of the 20 portfoliosβ value-weighted monthly returns on market excess
returns. Following Fama and French (1992), we use the entire sample to compute post-
ranking betas. That is, we pool together these six monthly returns across all calendar
months to estimate the portfolio beta. We estimate post-ranking betas for months 25-36
in a similar fashion. The two sets of post-ranking betas are then labelled π½11, ..., π½20
1 and
π½125, ..., π½20
25.
20 We sort stocks into vigintiles in order to increase the statistical precision of our cross-sectional estimate.
However, Appendix Table A4 confirms that our results are virtually identical if we instead sort stocks into
deciles.
20
To calculate the intercept and slope of the short-term and long-term security
market lines, we estimate the following cross-sectional regressions:
short-term SML: ππ ππ‘π,π‘1 = πππ‘ππππππ‘π‘
1 + π πππππ‘1π½π
1,
long-term SML: ππ ππ‘π,π‘25 = πππ‘ππππππ‘π‘
25 + π πππππ‘25π½π
25,
where ππ ππ‘π,π‘1 is portfolio πβs monthly excess returns in months 1 through 6, and ππ ππ‘π,π‘
25
is portfolio πβs monthly returns in months 25 through 36. These two regressions then give
us two time-series of coefficient estimates of the intercept and slope of the short-term and
long-term security market lines: ( πππ‘ππππππ‘π‘1, π πππππ‘
1 ) and ( πππ‘ππππππ‘π‘25, π πππππ‘
25 ),
respectively. As the average excess returns and post-ranking betas are always measured
at the same point in time, the pair (πππ‘ππππππ‘π‘1, π πππππ‘
1) fully describes the security market
line in the short run, while (πππ‘ππππππ‘π‘25, π πππππ‘
25) captures the security market line two
years later.
We then examine how these intercepts and slopes vary as a function of our measure
of beta-arbitrage capital. In particular, we conduct an OLS regression of the intercept and
slope measured in each month on lagged πΆππ΅π΄π . As can be seen from Table VI, the
intercept of the short-term security market line significantly increases in πΆππ΅π΄π , and its
slope significantly decreases in πΆππ΅π΄π . The top panel of Figure 3 shows this pattern
graphically. During high πΆππ΅π΄π βi.e., high beta-arbitrage capitalβperiods, the short-
term security market line strongly slopes downward, indicating strong profits to the low-
beta strategy, consistent with arbitrageurs expediting the correction of market
misevaluation. In contrast, during low πΆππ΅π΄π βi.e., low beta-arbitrage capitalβperiods,
the short-term security market line is weakly upward sloping and the beta-arbitrage
strategy, as a consequence, unprofitable, consistent with delayed correction of the beta
anomaly.
The pattern is completely reversed for the long-term security market line. The
intercept of the long-term security market line is significantly negatively related to πΆππ΅π΄π ,
whereas its slope is significantly positively related to πΆππ΅π΄π . As can be seen from the
21
bottom panel of Figure 3, two years after high πΆππ΅π΄π periods, the long-term security
market line turns upward sloping; indeed, the slope is so steep (resulting in a negative
intercept) that the beta strategy loses money, consistent with over-correction of the low
beta anomaly by crowded arbitrage trading. In contrast, after low πΆππ΅π΄π periods, the
long-term security market line turns downward sloping, reflecting eventual profitability
of the low-beta strategy in the long run.
IV.E. Smarter Beta-Arbitrage Strategies
One way to measure the economic importance of these boom and bust cycles is through
an out-of-sample calendar-time trading strategy. 21 We combine these time-varying
overreaction and subsequent reversal patterns as follows. We first time the standard beta-
arbitrage strategy using current πΆππ΅π΄π . If πΆππ΅π΄π is above the 80th percentile (of its
distribution up to that point), we are long the long-short beta-arbitrage strategy studied
in Table III for the next six months. Otherwise, we short that portfolio over that time
period. We skip the first three years of our sample to compute the initial distribution as
well as show in-sample results in Panel A of Table VII for the sake of comparison.
In addition, if πΆππ΅π΄π from three years ago is below the 20th percentile (of its prior
distribution), we are long for the next twelve months the long-short beta-arbitrage
strategy based on beta estimates from three years ago. Otherwise, we short that portfolio,
again for the next twelve months.
This strategy harvests beta-arbitrage profits much more wisely than unconditional
bets against beta. As can be seen from Panel B of Table VII, the four-factor alpha is 51
basis points per month with a t-statistic of 2.62. The six-factor alpha (where we add the
investment and profitability factors of Fama and French 2015a) remains high at 51 basis
points per month (t-statistic of 2.46). Finally, if we also include the BAB factor of Frazzini
21 This calendar-time approach also ensures that our Newey-West standard errors (for example, the standard
errors of Table III) are not misleading about the statistical significance of our findings.
22
and Pedersen (2014) as a seventh factor, the abnormal return increases to 63 basis points
per month with a t-statistic of 3.10. By comparison, the standard value-weight beta-
arbitrage strategy yields a four-factor alpha of -0.05% per month (t-statistic = -0.17) in
our sample period.
We have also estimated conditional regressions where we interact each factor with
πΆππ΅π΄π . The alpha from this regression is significantly larger at 0.87% per month (t-
statistic of 2.66).22
V. Testing the Economic Mechanism
The previous section documents rich cross-sectional and time-series variation in expected
returns linked to our proxy for arbitrage activity and the low-beta anomaly. In this
section, we delve deeper to test specific aspects of the economic mechanism we argue is
behind these patterns. Our interpretation of these patterns makes specific novel
predictions in terms of the role of firm leverage, the limits to arbitrage, and the reaction
of sophisticated investors to these patterns.
V.A. Beta Expansion
Beta arbitrage can be susceptible to positive-feedback trading. Successful bets on (against)
low-beta (high-beta) stocks result in prices of those securities rising (falling). If the
underlying firms are leveraged, this change in price will, all else equal, result in the
securityβs beta falling (increasing) further.23 Thus, not only do arbitrageurs not know when
to stop trading the low-beta strategy, their (collective) trades also affect the strength of
22 We construct an βeven-smarterβ beta arbitrage strategy by further exploiting differences between high-
leverage and low-leverage firms. In particular, we divide all stocks into four quartiles based on their lagged
leverage ratios. We then go long the smart-beta-strategy formed solely with high-leverage stocks and short
the smart-beta strategy solely with low-leverage stocks. This βeven-smarterβ beta strategy yields a monthly
alpha of 51bp (t-statistic = 2.93) after controlling for the Fama and French (2016) five factors, momentum
factor, BAB factor, as well as the smart-beta portfolio.
23 The idea that, all else equal, changes in leverage drive changes in equity beta is, of course, the key insight
behind Proposition II of Modigiliani and Miller (1958).
23
the signal. Consequently, beta arbitrageurs may increase their bets precisely when trading
becomes crowded and the expected profitability of the strategy has decreased.
We test this prediction in Table VIII. The dependent variable is the spread in betas
across the high and low value-weight beta decile portfolios, denoted π΅ππ‘πππππππ, as of
the end of year 1. The independent variables include lagged πΆππ΅π΄π , the beta-formation-
period value of π΅ππ‘πππππππ (computed from the same set of low- and high- beta stocks
as the dependent variable), the average book leverage quintile (πΏππ£πππππ) across the high
and low beta decile portfolios, and an interaction between πΆππ΅π΄π and πΏππ£πππππ. Note
that since we estimate beta using 52 weeks of stock returns, the two periods of beta
estimation that determine the change in π΅ππ‘πππππππ do not overlap. We include a trend
in all regressions, but our results are robust to not including the trend dummies.
Regression (1) in Table VIII shows that when πΆππ΅π΄π is relatively high, future
π΅ππ‘πππππππ is also high, controlling for lagged π΅ππ‘πππππππ. A one-standard-deviation
increase in πΆππ΅π΄π forecasts an increase in π΅ππ‘πππππππ of roughly 7% (of the average beta
spread). Regression (2) shows that this is particular true when πΏππ£πππππ is also high. If
beta-arbitrage bets were to contain the highest book-leverage quintile stocks, a one-
standard deviation increase in πΆππ΅π΄π would increase π΅ππ‘πππππππ by nearly 10%.
These results are consistent with a positive feedback channel for the beta-arbitrage
strategy that works through firm-level leverage. In terms of the economic magnitude of
this positive feedback loop, we draw a comparison with the price momentum strategy.
The formation-period spread for a standard price momentum bet in the post-1963 period
is around 115%, while the momentum profit in the subsequent year is close to 12% (e.g.,
Lou and Polk, 2014). Put differently, if we attribute price momentum entirely to positive
feedback trading, such trading increases the initial return spread by about 10% (12%
divided by 115%) in the subsequent year, which is similar in magnitude to the positive
feedback channel we document for beta arbitrage. Table VIII Panel B confirms that these
results are robust to the same methodological variations as in Table IV.
24
Table IX turns to firm-level regressions to document the beta expansion our story
predicts. In particular, we estimate panel regressions forecasting beta with lagged πΆππ΅π΄π .
At the end of each month, all stocks are sorted into deciles based on their market beta
calculated using daily returns in the past 12 months. The dependent variable is
πππ π‘ π ππππππ π΅ππ‘π, the change in stock beta from year t to t+1 (again, we use non-
overlapping periods). In addition to πΆππ΅π΄π , we also include π·ππ π‘ππππ, the difference
between a stockβs pre-formation beta and the average pre-formation beta in year t.
πΏππ£πππππ is the book leverage of the firm, measured in year t. We also include all double
and triple interaction terms of πΆππ΅π΄π , π·ππ π‘ππππ, and πΏππ£πππππ. Other control variables
include the lagged firm size, book-to-market ratio, lagged one-month and one-year stock
return, and the prior-year idiosyncratic volatility. Time-fixed effects are included in
Columns 3 and 4. Note that since πΆππ΅π΄π is a time-series variable, it is subsumed by the
time dummies in those regressions.
In all four regressions, stocks with higher Distance have a higher
πππ π‘ π ππππππ π΅ππ‘π, consistent with betas being persistent. This persistence is higher when
πΆππ΅π΄π is relatively high. Our main focus is on the triple interaction among πΆππ΅π΄π ,
π·ππ π‘ππππ, and πΏππ£πππππ. The persistence in a firmβs beta is significantly stronger when
πΆππ΅π΄π and πΏππ£πππππ are high. Taken together, these results are consistent with beta-
arbitrage activity causing the cross-sectional spread in betas to expand.
As a natural extension, our positive feedback channel suggests that booms and
busts of beta arbitrage should be especially strong among more highly levered stocks.
Figure 4 reports results where the sample is split based on leverage. Specifically, at the
beginning of the holding period, we sort stocks into four equal groups using book leverage.
For each leverage quartile, we compute the πΆππ΅π΄π return spread β i.e., the difference in
four-factor alpha to the beta arbitrage strategy between high and low πΆππ΅π΄π periods.
Reported in the figure is the cumulative difference in the πΆππ΅π΄π return spread between
the highest and lowest leverage quartiles over the five years after portfolio formation.
25
As can be seen from the figure, the difference in the πΆππ΅π΄π return spread rises
substantially in the first twelve months, by nearly 18% (1.46%*12, t-statistic = 2.14). It
then mostly reverses in the subsequent years. For example, the cumulative πΆππ΅π΄π return
spread in years 4-5 is roughly -10% (t-statistic = -2.41). This finding is consistent with
our novel positive feedback channel facilitating excessive arbitrage trading activity that
can potentially destabilize prices.
In results not reported, we have also confirmed that leverage splits enhance the
profitability of the calendar-time strategies studied in Table VII. Specifically, we go long
a version of the trading strategy restricted to the top quartile of firms based on leverage
and go short the corresponding low leverage (bottom quartile) version. The resulting in-
sample alpha is 51 basis points per month with a t-statistic of 2.93. The out-of-sample
estimate still generates a statistically-significant 36 basis points per month.
V.A.1. Conditional Attribution
Give that beta is moving with πΆππ΅π΄π , we also estimate conditional performance
attribution regressions (that is, we allow for the possibility that portfolio betas and
expected market and factor returns comove in the time series). The last rows of Table III
Panel A and Panel B report the results of those regressions. We find that the long-run
reversal of beta-arbitrage profits remains; there is again an economically large reversal of
beta-arbitrage profits when πΆππ΅π΄π is high. Figure 5 plots the conditional security market
line in the short and long-run as a function of lagged πΆππ΅π΄π . It is easy to see from the
figure our result that beta expansion and destabilization go hand-in-hand: the range of
average beta across the 20 beta-portfolios is much larger during high πΆππ΅π΄π periods than
in low πΆππ΅π΄π periods.
V.B. Low Limits to Arbitrage
26
We interpret our findings as consistent with arbitrage activity facilitating the correction
of the slope of the security market line in the short run. However, in periods of crowded
trading, arbitrageurs can cause price overshooting. In Table X, we exploit cross-sectional
heterogeneity to provide additional support for our interpretation. All else equal,
arbitrageurs prefer to trade stocks with low idiosyncratic volatility (to reduce tracking
error), high liquidity (to facilitate opening/closing of the position), and large capitalization
(to increase strategy capacity). As a consequence, we split our sample along each of these
dimensions. In particular, we rank stocks into quartiles based on the variable in question
(as of the beginning of the holding period); we label the quartile with the weakest limits
to arbitrage as βLow LTAβ and the quartile with the strongest limits to arbitrage as
βHigh LTA.β Our focus is on the long-run reversal associated with periods of high πΆππ΅π΄π .
The first two columns report results based on market capitalization, the third and
fourth based on idiosyncratic volatility, and the final two based on liquidity. The first
column of each pair shows the difference in four-factor alpha to the beta arbitrage strategy
between high πΆππ΅π΄π periods and low πΆππ΅π΄π periods in months 1-6 while the second
column shows the difference occurring in year 3.
πΆππ΅π΄π is an economically stronger predictor of abnormal returns to beta-arbitrage
strategies in months 1-6 for large stocks, stocks with low idiosyncratic volatility, and those
with high liquidity. Two of the estimates are statistically significant, and we can easily
reject the null hypothesis that the three estimates are jointly zero (the average value is
0.87 with a t-statistic of 2.73). For each of the three proxies for low limits to arbitrage,
we also find economically and statistically significant differences in the predictability of
year 3 returns. In summary, Table X confirms that our effect is stronger among stocks
with weaker limits of arbitrage, exactly where one expects arbitrageurs to play a more
important role.
V.C. Time-series and Cross-sectional Variation in Fund Exposures
27
We next use our novel measure of beta-arbitrage activity to understand time-series and
cross-sectional variation in the performance of long-short/market-neutral hedge funds,
typically considered to be the classic example of smart money; as well as active mutual
funds, who are subject to more stringent leverage and short-sale constraints. Table XI
reports estimates of panel regressions of monthly fund returns on the Fama-French-
Carhart four-factor model augmented with the beta-arbitrage factor of Frazzini and
Pedersen (2014). In particular, we allow the coefficient on the Frazzini-Pedersen betting-
against-beta (BAB) factor to vary as a function of πΆππ΅π΄π , a fundβs AUM, and the
interaction between these two variables. To capture variation in a fundβs AUM, we create
a dummy-variable, πππ§ππ πππ, that takes the value of zero if the fund is in the smallest-
AUM tercile (within the active mutual fund or long-short equity hedge fund industry,
depending on the returns being analyzed) in the previous month, one if it is in the middle
tercile, and two otherwise. The first two columns analyze hedge fund returns while the
last two columns analyze active mutual fund returns.
We find that the typical long-short equity hedge fund increases their exposure to
the BAB factor when πΆππ΅π΄π is relatively high. For the 20% of the sample period that is
associated with the lowest values of πΆππ΅π΄π , the typical hedge fundβs BAB loading is
-0.08. This loading increases by 0.022 for each increment in πΆππ΅π΄π rank. (It is noteworthy
that the average long-short hedge fund is loading negatively on the BAB factor β i.e., on
average, funds are tilting towards high beta stocks.)
Adding the interaction with AUM reveals that the ability of hedge funds to time
beta-arbitrage strategies is decreasing in the size of the fundβs assets under management.
These findings seem reasonable as we would expect large funds to be unable to time a
beta-arbitrage strategy as easily as smaller (and presumably nimbler) funds.
The typical small fundβs exposure increases by 0.035 for each increase in πΆππ΅π΄π
rank. Thus, when πΆππ΅π΄π is in the top quintile, the typical small hedge fundβs BAB loading
is 0.048. In contrast, large hedge fundsβ BAB loading moves from -0.118 to only 0.022
28
across the five πΆππ΅π΄π quintiles, a much smaller increase in exposure to beta arbitrage.
Indeed, when πΆππ΅π΄π is high, small hedge funds have loadings on BAB that are more than
twice as large.
As can be seen from columns 3 and 4, there is a vastly different pattern in the
market exposures of mutual funds. To start, mutual funds have an average market beta
that is larger than one, and hence unsurprisingly load negatively on the BAB factor. This
suggests that mutual funds on average overweight high beta stocks β consistent with the
intuition that most mutual funds are leverage-constrained. Second, none of the
interactions are statistically significant. In particular, mutual fundsβ loadings on market
risk do not vary with πΆππ΅π΄π , our proxy for the strategyβs crowdedness.
V.D. Fresh versus Stale Beta
Though beta-arbitrage activity may cause the beta spread to vary through time, for a
feedback loop to occur, beta arbitrageurs must base their strategies on fresh estimates of
beta rather than on stale estimates. (Note that the autocorrelation in a stockβs market
beta is far less than one.) Consistent with this claim, we show that our predictability
results decay as a function of beta staleness.
We repeat the previous analysis of section IV.B, but replacing our fresh beta
estimates (measured over the most recent year) with progressively staler ones. In
particular, we estimate betas in each of the five years prior to the formation year. As a
consequence, both the resulting beta strategy and the associated πΆππ΅π΄π are different for
each degree of beta staleness. For each of these six beta strategies, we regress the four-
factor alpha of the strategy in months one-six and year three on its corresponding πΆππ΅π΄π .
Figure 6 plots the resulting regression coefficients (results for months 1-6 plotted
with a blue square and results for year 3 plotted with a red circle) as a function of the
degree of staleness of beta. The baseline results with the most recent beta are simply the
corresponding figures from Table V. We find that both the short-run and long-run
29
predictability documented in section IV.B decay as the beta signal becomes more and
more stale. Indeed, strategies using beta estimates that are five years old display no
predictability. These results are consistent with the feedback loop we propose.24
VI. Conclusion
We study the response of arbitrageurs to a flat security market line. Using an approach
to measuring arbitrage activity first introduced by Lou and Polk (2014), we document
booms and busts in beta arbitrage. Specifically, we find that when arbitrage activity is
relatively low, abnormal returns on beta-arbitrage strategies take much longer to
materialize, appearing only two to three years after putting on the trade. In sharp
contrast, when arbitrage activity is relatively high, abnormal returns on beta-arbitrage
strategies occur relatively quickly, within the first six months of the trade. These strong
abnormal returns then revert over the next three years. Thus, our findings are consistent
with arbitrageurs exacerbating this time-variation in the expected return to beta
arbitrage.
We provide evidence on a novel positive feedback channel for beta-arbitrage
activity. Since the typical firm is levered and given the mechanical effects of leverage on
equity beta (Modigliani and Miller 1958), buying low-beta stocks and selling high-beta
stocks may cause the cross-sectional spread in betas to increase. We show that this beta
expansion occurs when beta-arbitrage activity is high and particularly so when stocks
typically traded by beta arbitrageurs are highly levered. Thus, beta arbitrageurs may
actually increase their bets when the profitability of the strategy has decreased. Indeed,
we find that the short-run abnormal returns to high-leverage beta-arbitrage stocks more
than triples before reverting in the long run.
24 Figure A1 documents that the unconditional CAPM alpha for a beta-arbitrage strategy also significantly
declines with beta staleness.
30
Interestingly, the unconditional four-factor alpha of a value-weight beta-arbitrage
strategy over our 1970-2010 sample is close to zero, much lower than the positive value
one finds for earlier samples. Thus, it seems that arbitrageursβ response to Black, Jensen,
and Scholesβs (1972) famous finding has been right on average. However, our conditional
analysis reveals rich time-series variation that is consistent with the general message of
Stein (2009): Arbitrage activity faces a significant coordination problem for unanchored
strategies that have positive feedback characteristics.
31
References
Amihud, Yakov, 2002, Illiquidity and stock returns: Cross-section and time series effects, Journal of Financial Markets 5, 31β56.
Antoniou, Constantinos, John A. Doukas, and Avanidhar Subrahmanyam, 2015, Investor Sentiment, Beta, and the Cost of Equity Capital, Management Science 62, 347-367.
Anton, Miguel and Christopher Polk, 2014, Connected Stocks, Journal of Finance 69,
1099-1127.
Asness, Clifford, Jacques Friedman, Robert Krail, and John Liew, 2000, Style Timing:
Value versus Growth, Journal of Portfolio Management 26, 50-60.
Baker, Malcolm, Brendan Bradley and Jeffrey Wurgler, 2011, Benchmarks as Limits to
Arbitrage: Understanding the Low-Volatility Anomaly, Financial Analysts Journal 67, 40-54.
Baker, Malcolm, Brendan Bradley, and Ryan Taliaferro. 2014, The Low Risk Anomaly:
A Decomposition into Micro and Macro Effects, Financial Analysts Journal 70, 43β58.
Baker, Malcolm, and Jeffrey Wurgler, 2006, Investor Sentiment and the Cross-section of
Stock Returns, Journal of Finance 61, 1645
Baker, Malcolm, and Jeffry Wurgler, 2007, Investor sentiment in the stock market,
Journal of Economic Perspectives 21, 129-152.
Barberis, Nicholas and Ming Huang, Stocks as Lotteries: The Implications of Probability Weighting for Security Prices, American Economic Review 98, 2066-2100.
Barberis, Nicholas and Andrei Shleifer, 2003, Style investing, Journal of Financial Economics 68, 161-199.
Barberis, Nicholas, Andrei Shleifer, and Jeffrey Wurgler, 2005, Comovement, Journal of Financial Economics 75, 283-317.
Black, Fischer. 1972, Capital Market Equilibrium with Restricted Borrowing, Journal of Business. 45, 444-454.
Black, Fischer, Michael C. Jensen and Myron Scholes. 1972. βThe Capital Asset Pricing
Model: Some Empirical Tests,β in Studies in the Theory of Capital Markets. Michael C. Jensen, ed. New York: Praeger, pp. 79-121.
Blitz, David, and P. V. Vliet, 2007, The Volatility Effect: Lower Risk Without Lower
Return, Journal of Portfolio Management 34, 102-113.
32
Blitz, David, J. Pang, P. V. Vliet, 2013, The Volatility Effect in Emerging Markets, Emerging Markets Review 16, 31-45.
Buffa, Andrea, Dimitri Vayanos, and Paul Woolley, 2014, Asset Management Contracts
and Equilibrium Prices, London School of Economics working paper.
Campbell, John, Stefano Giglio, Christopher Polk, and Robert Turley, 2018, An
Intertemporal CAPM with Stochastic Volatility, Journal of Financial Economics, forthcoming.
Cao, Charles, Yong Chen, Bing Liang, and Andrew Lo, Can Hedge Funds Time Market Liquidity, 2013, Journal of Financial Economics 109, 493-516.
Carhart, Mark, 1997, On Persistence in Mutual Fund Performance, Journal of Finance 52, 56β82.
Cohen, Randolph B., Christopher Polk, and Tuomo Vuolteenaho, 2005, The Value Spread, Journal of Finance 58, 609-641.
Cohen, Randolph B., Christopher Polk, and Tuomo Vuolteenaho, 2005, Money Illusion in
the Stock Market: The Modigliani-Cohn Hypothesis, Quarterly Journal of Economics, 70,
639-668.
Daniel, Kent, Mark Grinblatt, Sheridan Titman, and Russ Wermers, 1997, Measuring Mutual Fund Performance with Characteristic Based Benchmarks, Journal of Finance 52,
1035-1058.
DeLong, J. Bradford, Andrei Shleifer, Lawrence H. Summers, and Robert Waldmann,
1990, Positive Feedback Investment Strategies and Destabilizing Rational Speculation, Journal of Finance 45, 379-395.
Fama, Eugene F., 1970, Efficient Capital Markets: A Review of Theory and Empirical Work, Journal of Finance 25, 383-417.
Fama, Eugene F. and Kenneth R. French, 1992, The Cross-Section of Expected Stock
Returns, Journal of Finance, 47, 427-465.
Fama, Eugene F. and Kenneth R. French, 1993, Common Risk Factors in the Returns on
Stocks and Bonds, Journal of Financial Economics 33, 3-56.
Fama, Eugene F. and Kenneth R. French, 1996, The CAPM is Wanted, Dead or Alive,
Journal of Finance 51, 1947-1958.
Fama, E. F., and K. R. French. 2015a. A five-factor asset pricing model, Journal of Financial Economics 116: 1-22.
33
Fama, E. F., and K. R. French. 2015b. Dissecting anomalies with a five-factor model, Review of Financial Studies, forthcoming.
Frazzini, Andrea and Lasse H. Pedersen, 2014, Betting Against Beta, Journal of Financial Economics 111, 1-25.
Greenwood, Robin and David Thesmar, 2011, Stock Price Fragility, Journal of Financial Economics 102, 471β490.
Hanson, Samuel G. and Adi Sunderam, The Growth and Limits of Arbitrage: Evidence
from Short Interest" Review of Financial Studies 27, 1238β1286.
Hong, Harrison and David Sraer, 2016, Speculative Betas, Journal of Finance 71, 2095-
2144.
JylhΓ€, Petri, 2017, Margin Requirements and the Security Market Line, Journal of Finance, forthcoming.
J.P. Morgan, 2014, ADR Arbitrage and Program Balances, DR Advisor Insights.
Karceski, Jason, 2002, Returns-Chasing Behavior, Mutual Funds, and Betaβs Death, Journal of Financial and Quantitative Analysis 37, 559-594.
Koijen, Ralph and Motohiro Yogo, 2017, An Equilibrium Model of Institutional Demand
and Asset prices, NBER working paper 21749.
Kumar, Alok, 2009, Who gambles in the stock market?, Journal of Finance 64, 1889β1933.
Lintner, John, 1965, The Valuation of Risky Assets and the Selection of Risky Investments
in Stock Portfolios and Capital Budgets, Review of Economics and Statistics 47, 13-37.
Liu, Jianan, Robert F. Stambaugh, and Yu Yuan, 2017, Absolving Beta of Volatilityβs
Effects, Journal of Financial Economics, Forthcoming.
Lou, Dong, 2012, A Flow-Based Explanation for Return Predictability, Review of Financial Studies 25, 3457-3489.
Lou, Dong, and Christopher Polk, 2014, Comomentum: Inferring Arbitrage Activity from
Return Correlations, London School of Economics working paper.
Mehrling, Perry, 2005, Fischer Black and The Revolutionary Idea of Finance. Hoboken, NJ: John Wiley and Sons.
Miller, Edward M., 1977, Risk, Uncertainty, and Divergence of Opinion, Journal of Finance 32, 1151-1168.
34
Modigliani, Franco, and Merton H. Miller, 1958, The cost of capital, corporation finance, and the theory of investment, American Economic Review 48, 655β669.
Modigliani, Franco, and Richard Cohn, 1979, Inflation, Rational Valuation, and the
Market, Financial Analysts Journal 35, 24β44.
PΓ‘stor, LuboΕ‘ and Robert Stambaugh, 2003, Liquidity risk and expected stock returns,
Journal of Political Economy 111, 642β685.
Pontiff, Jeff and R. David McLean, 2016, Does Academic Publication Destroy Stock
Return Predictability, Journal of Finance 71, 5-32.
Pollet, Joshua and Mungo Wilson, 2010, Average Correlation and Stock Market
Returns, Journal of Financial Economics 96, 364-380.
Polk, Christopher, Sam Thompson, and Tuomo Vuolteenaho, 2006, Cross-sectional forecasts of the equity premium, Journal of Financial Economics 81, 101-141.
RΓΆsch, Dominik, 2014, The impact of arbitrage on market liquidity, Erasmus University
working paper.
Schwert, G. William, 2003, Anomalies and Market Efficiency, Chapter 15 in Handbook of the Economics of Finance, eds. George Constantinides, Milton Harris, and Rene M. Stulz, North-Holland 937-972.
Sharpe, William, 1964, Capital Asset Prices: A Theory of Market Equilibrium under Conditions of Risk, Journal of Finance 19, 425-442.
Shleifer, Andrei and Robert W. Vishny, 1997, The Limits of Arbitrage, Journal of Finance 52, 35β55.
Stambaugh, Robert F., Yu, Jianfeng, Yuan, Yu, 2015. Arbitrage asymmetry and the
idiosyncratic volatility puzzle, Journal of Finance 70, 1903β1948.
Stein, Jeremy, 2009, Sophisticated Investors and Market Efficiency, Journal of Finance 64, 1517-1548.
Table I: Summary Statistics
This table provides characteristics of βπΆππ΅π΄π ,β the excess comovement among low beta stocks over the period
1970-2013 (we then examine beta arbitrage returns in the following three years). At the end of each month,
all stocks are sorted into deciles based on their market beta calculated using daily returns in the past 12
months. To account for illiquidity and non-synchronous trading, we include on the right-hand side of the
regression equation five lags of the excess market return, in addition to the contemporaneous excess market
return. The pre-ranking beta is simply the sum of the six coefficients from the OLS regression. Pairwise
partial return correlations (controlling for the Fama-French three factors) for all stocks in the bottom beta
decile are computed based on weekly stock returns in the previous 12 months. πΆππ΅π΄π is the average pair-
wise correlation between any two stocks in the low-beta decile in year π‘. πΌπππππ‘πππ is the smoothed inflation
rate used by Cohen, Polk, and Vuolteenaho (2005) who apply an exponentially weighted moving average
(with a half-life of 36 months) to past log growth rates of the producer price index. ππππ‘πππππ‘ is the sentiment
index proposed by Wurgler and Baker (2006, 2007). π·ππ πππππππππ‘ is the beta-weighted standard deviation
of analystsβ long-term growth rate forecasts, as used in Hong and Sraer (2016). πππ ππππππ is the difference
between the LIBOR rate and the US Treasury bill rate. Panel A reports the summary statistics of these
variables. Panel B shows the time-series correlations among these variables for the entire sample period.
Panel A: Summary Statistics
Variable N Mean Std. Dev. Min Max
πΆππ΅π΄π 528 0.105 0.026 0.037 0.203
Inflation 528 0.003 0.002 0.000 0.008
Sentiment 528 0.012 0.941 -2.325 3.076
Disagreement 384 -0.013 1.013 -1.277 3.593
Ted Spread 336 0.621 0.438 0.118 3.353
Panel B: Correlation
πΆππ΅π΄π Inflation Sentiment Disagreement Ted Spread
πΆππ΅π΄π 1
Inflation -0.324 1
Sentiment 0.013 -0.394 1
Disagreement 0.336 -0.162 0.187 1
Ted Spread 0.267 0.241 -0.034 -0.181 1
Table II: Determinants of πΆππ΅π΄π
This table reports regressions of πΆππ΅π΄π , described in Table I, on lagged variables plausibly linked to arbitrage
activity in the post-1994 period (constrained by the availability of the hedge fund AUM data). At the end of
each month, all stocks are sorted into deciles based on their market beta calculated using daily returns in the
past 12 months. To account for illiquidity and non-synchronous trading, we include on the right-hand side of
the regression equation five lags of the excess market return, in addition to the contemporaneous excess
market return. The pre-ranking beta is simply the sum of the six coefficients from the OLS regression. The
dependent variable in the regressions, πΆππ΅π΄π , is the average pairwise partial weekly return correlation in the
low-beta decile over 12 months. πΌππ π‘ ππ€π is the aggregate institutional ownership of the low-beta decile,
π΄ππ is the logarithm of the total assets under management of long-short equity hedge funds (detrended).
π΅π΄π΅ π΄ππβπ is the realized four-factor alpha of Frazzini and Pedersenβs BAB factor. πΌπππππ‘πππ is the smoothed
inflation rate used by Cohen, Polk, and Vuolteenaho (2005) who apply an exponentially weighted moving
average (with a half-life of 36 months) to past log growth rates of the producer price index. ππππ‘πππππ‘ is the
sentiment index proposed by Wurgler and Baker (2006, 2007). π·ππ πππππππππ‘ is the beta-weighted standard
deviation of analystsβ long-term growth rate forecasts, as used in Hong and Sraer (2012). πππ ππππππ is the
difference between the LIBOR rate and the US Treasury bill rate. We also include in the regression the
Pastor-Stambaugh liquidity factor (PS Liquidity). A trend dummy is included in all regression specifications.
All independent variables are divided by their corresponding standard deviation, so that the coefficient
represents the effect of a one-standard-deviation change in the independent variable on πΆππ΅π΄π . Standard
errors are shown in brackets. *, **, *** denote significance at the 10%, 5%, and 1% level, respectively.
DepVar πΆππ΅π΄π π‘
[1] [2]
πΌππ π‘ ππ€ππ‘β1 0.026*** 0.006**
[0.003] [0.003]
π΄πππ‘β1 0.004** 0.003**
[0.002] [0.002]
π΅π΄π΅ π΄ππβππ‘β1 0.000 0.004***
[0.001] [0.001]
πΌπππππ‘ππππ‘ -0.005
[0.003]
ππππ‘πππππ‘π‘ 0.003
[0.002]
π·ππ πππππππππ‘π‘ 0.011***
[0.002]
πππ πππππππ‘ 0.016***
[0.002]
ππ πΏπππ’ππππ‘π¦π‘ 0.006*** 0.010***
[0.002] [0.001]
TREND Yes Yes
Adj-R2 0.350 0.529
No. Obs. 228 228
Table III: Forecasting Beta-arbitrage Returns with πΆππ΅π΄π
This table reports returns to the beta arbitrage strategy as a function of lagged πΆππ΅π΄π . At the end of each
month, all stocks are sorted into deciles based on their market beta calculated using daily returns in the past
12 months. To account for illiquidity and non-synchronous trading, we include on the right-hand side of the
regression equation five lags of the excess market return, in addition to the contemporaneous excess market
return. The pre-ranking beta is simply the sum of the six coefficients from the OLS regression. All months
are then classified into five groups based on πΆππ΅π΄π , the average pairwise partial weekly return correlation in
the low-beta decile over the past 12 months. Reported below are the returns to the beta arbitrage strategy
(i.e., to go long the value-weight low-beta decile and short the value-weighted high-beta decile) in each of the
three years after portfolio formation during 1970 to 2013, following low to high πΆππ΅π΄π . Panels A and B
report, respectively, the average monthly three-factor alpha and Carhart four-factor alpha of the beta
arbitrage strategy. β5-1β is the difference in monthly returns to the long-short strategy following high vs. low
πΆππ΅π΄π ; β5-1 Conditionalβ is the difference in conditional abnormal returns (i.e., allowing for risk loadings to
vary as a function of πΆππ΅π΄π ) following high vs. low πΆππ΅π΄π . T-statistics, shown in parentheses, are computed
based on standard errors corrected for serial-dependence up to 24 lags. 5% statistical significance is indicated
in bold.
Panel A: Fama-French-Adjusted Beta-arbitrage Returns
Months 1-6 Months 7-12 Year 2 Year 3 Years 2&3
Rank Obs. Est. t-stat Est. t-stat Est. t-stat Est. t-stat Est. t-stat
1 105 0.58% (2.20) 0.58% (2.41) 0.65% (2.40) 0.88% (3.41) 0.73% (2.72)
2 106 -0.16% (-0.53) 0.56% (1.91) 0.74% (2.70) 0.17% (0.82) 0.49% (2.36)
3 106 -0.35% (-1.19) 0.20% (0.76) 0.41% (1.57) 0.55% (2.29) 0.49% (2.24)
4 106 -0.19% (-0.69) 0.33% (1.18) -0.09% (-0.28) -0.01% (-0.02) -0.03% (-0.11)
5 105 1.47% (3.62) 0.37% (0.81) 0.02% (0.04) -0.60% (-1.68) -0.28% (-1.59)
5-1 0.89% (1.89) -0.21% (-0.41) -0.63% (-1.10) -1.48% (-3.17) -1.01% (-2.66)
5-1(Cond) 0.30% (0.74) -0.62% (-1.32) -0.85% (-1.54) -1.34% (-3.06) -1.05% (-3.08)
Panel B: Four-Factor Adjusted Beta-arbitrage Returns
Months 1-6 Months 7-12 Year 2 Year 3 Years 2&3
Rank Obs. Est. t-stat Est. t-stat Est. t-stat Est. t-stat Est. t-stat
1 105 0.33% (1.21) 0.29% (1.17) 0.40% (1.49) 0.59% (2.22) 0.45% (1.67)
2 106 -0.36% (-1.28) 0.19% (0.64) 0.46% (1.69) -0.03% (-0.13) 0.25% (1.06)
3 106 -0.57% (-1.97) -0.08% (-0.32) 0.08% (0.27) 0.30% (1.19) 0.21% (0.81)
4 106 -0.49% (-1.79) -0.01% (-0.03) -0.41% (-1.10) -0.35% (-0.85) -0.36% (-0.98)
5 105 1.08% (2.63) 0.14% (0.33) -0.14% (-0.31) -0.93% (-2.66) -0.53% (-2.58)
5-1 0.75% (1.57) -0.15% (-0.30) -0.54% (-1.02) -1.52% (-3.33) -0.99% (-2.52)
5-1 (Cond) 0.08% (0.20) -0.74% (-1.63) -1.02% (-2.15) -1.68% (-4.08) -1.33% (-3.89)
Table IV: Robustness Checks
This table reports returns to the beta arbitrage strategy as a function of lagged πΆππ΅π΄π . At the end of each
month, all stocks are sorted into deciles based on their market beta calculated using daily returns in the past
12 months. To account for illiquidity and non-synchronous trading, we include on the right-hand side of the
regression equation five lags of the excess market return, in addition to the contemporaneous excess market
return. The pre-ranking beta is simply the sum of the six coefficients from the OLS regression. All months
are then classified into five groups based on πΆππ΅π΄π , the average pairwise partial weekly return correlation in
the low-beta decile over the past 12 months. Reported below is the difference in four-factor alpha to the beta
arbitrage strategy between high πΆππ΅π΄π periods and low πΆππ΅π΄π periods. Year zero is the beta portfolio
ranking period. Row 1 shows the baseline results which are also reported in Table III. Row 2 shows the same
analysis for the earlier sample (1927-1969) as a placebo test. In rows 3 and 4, we exclude the tech bubble
crash and the recent financial crisis from our sample. In rows 5-8, we rank all months based on smoothed
past inflation (Cohen, Polk, and Vuolteenaho, 2005), sentiment index (Wurgler and Baker, 2006), aggregate
analyst forecast dispersion (Hong and Sraer, 2014), and Ted Spread (Frazzini and Pedersen, 2014),
respectively. In row 9, we also control for the UMD factor in computing πΆππ΅π΄π . In row 10, we control for
both large- and small-cap HML in computing πΆππ΅π΄π . In row 11, we control for the Fama-French five factor
model that adds profitability and investment to their three-factor model. In row 12, we perform the entire
analysis on an industry-adjusted basis by sorting stocks into beta deciles within industries. In row 13, we
instead measure the correlation between the high and low-beta portfolios, with a low correlation indicating
high arbitrage activity. In Rows 14 and 15, we examine upside and downside πΆππ΅π΄π , as distinguished by the
median low-beta portfolio return. In Rows 16-22, we replace πΆππ΅π΄π with residual πΆππ΅π΄π from a time-series
regression where we purge from πΆππ΅π΄π variation linked to, respectively, πΆππππ and πΆπππππ’π (Lou and
Polk, 2017), the average pair-wise correlation in the market, the lagged 36-month volatility of the BAB factor
(Frazzini and Pedersen, 2013), market volatility over the past 24 months, a trend, and lagged πΆππ΅π΄π (where
we hold the stocks in the low-beta decile constant but calculate πΆππ΅π΄π using returns from the previous year).
In Row 23, abnormal returns to the beta arbitrage strategy is calculated relative to the Fama-French 5 factor
model plus the momentum factor. T-statistics, shown in parentheses, are computed based on standard errors
corrected for serial-dependence up to 24 lags. 5% statistical significance is indicated in bold.
Four-Factor Adjusted Beta-arbitrage Returns
Year 3 Years 2-3
Estimate t-stat Estimate t-stat
Subsamples
Full Sample: 1970-2010 -1.52% (-3.33) -0.99% (-2.52)
Early sample: 1927-1969 0.19% (0.41) 0.22% (0.68)
Excluding 2001 -1.56% (-3.44) -0.99% (-2.24)
Excluding 2007-2009 -1.48% (-3.06) -0.94% (-2.29)
Other predictors of beta-arbitrage returns
Inflation 0.13% (0.24) 0.27% (0.54)
Sentiment -0.11% (-0.17) 0.07% (0.11)
Disagreement 0.67% (1.03) 0.86% (1.83)
Ted Spread -0.21% (-0.41) -0.18% (-0.45)
Alternative definitions of πΆππ΅π΄π
Controlling for UMD -1.59% (-3.70) -1.09% (-2.65)
Controlling for Large/Small-Cap HML -1.53% (-3.35) -0.97% (-2.47)
Controlling for FF Five Factors -1.32% (-2.67) -1.00% (-2.54)
Controlling for Industry Factors -1.29% (-3.02) -0.93% (-2.37)
Correl btw High and Low Beta Stocks -0.88% (-2.33) -0.81% (-2.19)
Upside πΆππ΅π΄π -0.99% (-2.29) -0.83% (-2.26)
Downside πΆππ΅π΄π -0.30% (-0.67) 0.23% (0.94)
Residual πΆππ΅π΄π
Controlling for CoMomentum -1.65% (-3.78) -1.10% (-2.91)
Controlling for CoValue -1.57% (-3.63) -1.00% (-2.52)
Controlling for MKT CORR -1.82% (-4.21) -1.21% (-2.93)
Controlling for Vol(BAB) -1.66% (-3.81) -1.07% (-2.74)
Controlling for Mktvol24 -1.50% (-3.24) -0.98% (-2.57)
Controlling for Trend -1.51% (-3.31) -0.98% (-2.49)
Controlling for Pre-formation CoBAR -1.56% (-3.27) -1.13% (-2.85)
Beta arbitrage alpha based on
Controlling for FF5+MOM -1.46% (-3.52) -1.03% (-2.99)
Table V: Regression Analysis
This table reports returns to the beta arbitrage strategy as a function of lagged πΆππ΅π΄π . At the end of each
month, all stocks are sorted into deciles based on their market beta calculated using daily returns in the past
12 months. To account for illiquidity and non-synchronous trading, we include on the right-hand side of the
regression equation five lags of the excess market return, in addition to the contemporaneous excess market
return. The pre-ranking beta is simply the sum of the six coefficients from the OLS regression. The dependent
variable is the four-factor alpha of the beta arbitrage strategy (i.e., a portfolio that is long the value-weight
low-beta decile and short the value-weighted high-beta decile). The main independent variable is πΆππ΅π΄π , the
average pairwise partial weekly three-factor residual correlation within the low-beta decile over the past 12
months. We also include in the regression exponentially weighted moving average past inflation (Cohen, Polk,
and Vuolteenaho, 2005), a sentiment index (Wurgler and Baker, 2006), aggregate analyst forecast dispersion
(Hong and Sraer, 2016), Ted Spreadβthe difference between the LIBOR rate and the US Treasury bill rate,
the ValueSpreadβthe spread in log book-to-market-ratios across the low-beta and high-beta deciles, and
market volatility over the past 24 months. The first three columns examine returns to the beta arbitrage
strategy in months 1-6, and the next three columns examine the returns in year 3 after portfolio formation.
We report results based on Carhart four-factor adjustments. T-statistics, shown in brackets, are computed
based on standard errors corrected for serial-dependence up to 12 lags. *, **, *** denote significance at the
10%, 5%, and 1% level, respectively.
DepVar Four-Factor Alpha to the Beta Arbitrage Strategy
Months 1-6 Year 3
[1] [2] [3] [4] [5] [6]
πΆππ΅π΄π 0.145** 0.177*** 0.152* -0.195*** -0.154*** -0.277*** [0.067] [0.072] [0.087] [0.054] [0.053] [0.079]
πΌπππππ‘πππ 1.384 3.316 1.685* 2.793 [1.195] [2.253] [0.962] [3.358]
ππππ‘πππππ‘ 0.004** -0.003 0.002 0.002 [0.002] [0.003] [0.002] [0.004]
π·ππ πππππππππ‘ 0.006*** 0.005
[0.003] [0.003]
πππ ππππππ -0.008 0.001
[0.007] [0.008]
ππππ’πππππππ 0.002 0.004 0.001 -0.002
[0.005] [0.004] [0.003] [0.003]
πππ‘π£ππ24 0.065 -0.092 0.118 0.045
[0.106] [0.148] [0.150] [0.214]
Adj-R2 0.024 0.048 0.085 0.094 0.128 0.154
No. Obs. 528 528 336 528 528 336
Table VI: Predicting the Security Market Line
This table reports regressions of the intercept and slope of the security market line on lagged πΆππ΅π΄π . At the
end of each month, all stocks are sorted into vigintiles based on their market beta calculated using daily
returns in the past 12 months. To account for illiquidity and non-synchronous trading, we include on the
right-hand side of the regression equation five lags of the excess market return, in addition to the
contemporaneous excess market return. The pre-ranking beta is simply the sum of the six coefficients from
the OLS regression. We then estimate security market lines using monthly portfolio returns in months 1-6,
based on these 20 portfolios formed in each period. The post-ranking betas are calculated by regressing each
of the 20 portfoliosβ value-weighted monthly returns on the corresponding market return. Following Fama
and French (1992), we use the entire sample to compute post-ranking betas. The dependent variable in Panel
A is the intercept of the SML, while that in Panel B is the slope of the SML. The main independent variable
is πΆππ΅π΄π , the average pairwise partial weekly three-factor residual correlation within the low-beta decile over
the past 12 months. We also include in the regressions smoothed inflation, sentiment index, aggregate analyst
forecast dispersion, and Ted Spread. Other (unreported) control variables include the contemporaneous
market excess return, SMB return, and HML return. Standard errors, shown in brackets, are computed based
on standard errors corrected for serial-dependence. *, **, *** denote significance at the 10%, 5%, and 1%
level, respectively.
Panel A: DepVar = Intercept of SML
Months 1-6 Year3
πΆππ΅π΄π 0.171*** 0.226*** 0.134** -0.266*** -0.189*** -0.226*** [0.070] [0.051] [0.065] [0.077] [0.067] [0.087]
πΌπππππ‘πππ 2.585*** 5.728*** 0.422 1.069 [0.768] [1.670] [1.229] [2.157]
ππππ‘πππππ‘ 0.006*** 0.004 -0.001 -0.004 [0.001] [0.003] [0.002] [0.004]
π·ππ πππππππππ‘ 0.004*** 0.001 [0.002] [0.003]
πππ ππππππ -0.011*** 0.004 [0.004] [0.004]
Adj-R2 0.044 0.366 0.511 0.119 0.371 0.471
No. Obs. 528 528 336 528 528 336
Panel B: DepVar = Slope of SML
Months 1-6 Year3
πΆππ΅π΄π -0.337*** -0.215*** -0.130** 0.289*** 0.207*** 0.203** [0.088] [0.050] [0.066] [0.096] [0.069] [0.094]
πΌπππππ‘πππ -2.465*** -5.716*** -0.833 -0.006 [0.792] [1.706] [1.191] [2.583]
ππππ‘πππππ‘ -0.006*** -0.005* 0.001 0.005 [0.001] [0.003] [0.002] [0.004]
π·ππ πππππππππ‘ -0.003* 0.000 [0.002] [0.004]
πππ ππππππ 0.012*** -0.003 [0.004] [0.004]
Adj-R2 0.092 0.668 0.732 0.104 0.468 0.498
No. Obs. 528 528 336 528 528 336
Table VII: Smarter Beta-Arbitrage Strategies
This Table reports monthly returns to a βsmarterβ beta-arbitrage strategy that exploits the time-varying
overreaction and subsequent reversal in standard beta arbitrage strategies. Specifically, we first time the
standard beta-arbitrage strategy using current πͺππ©π¨πΉ . If πͺππ©π¨πΉ is above the 80th percentile, we go long
the long-short beta-arbitrage strategy studied in Table III for the next six months. Otherwise, we short that
portfolio over that time period. In addition, if πͺππ©π¨πΉ from three years ago is below the 20th percentile, we
are long for the next twelve months the long-short beta-arbitrage strategy based on beta estimates from three
years ago. Otherwise, we short that portfolio, again for the next twelve months. In Panel A, the percentile
break points of the πͺππ©π¨πΉ distribution are identified using the entire πͺππ©π¨πΉ time series (thus an in-sample
analysis). In Panel B, the percentile break points are identify based solely on its prior distribution (an out-
of-sample analysis); we skip the first three years of our sample to compute the initial distribution. In the full
specification of risk adjustment, we use a seven-factor model that includes: the Fama and French (2015) five-
factors (market, size, value, investment and profitability), the Carhart momentum factor, and Frazzini and
Pedersenβs (2014) betting-against beta(BAB) factor. 5% statistical significance is indicated in bold.
ALPHA RM-RF SMB HML UMD RMW CMA BAB
Panel A: In-Sample
0.53% 0.52 0.10 0.04 -0.15
(2.65) (10.62) (1.52) (0.37) (-2.07)
0.49% 0.51 0.19 0.06 -0.15 0.16 -0.12
(2.34) (10.38) (2.43) (0.55) (-2.10) (1.20) (-0.58)
0.62% 0.56 0.26 0.20 -0.08 0.39 0.02 -0.45
(2.98) (11.69) (3.38) (1.62) (-1.10) (2.96) (0.10) (-5.16)
Panel B: Out-of-Sample
0.51% 0.52 0.10 -0.02 -0.20
(2.62) (11.16) (1.51) (-0.25) (-2.86)
0.51% 0.51 0.17 0.04 -0.19 0.10 -0.19
(2.46) (10.64) (2.14) (0.36) (-2.76) (0.66) (-0.95)
0.63% 0.55 0.24 0.18 -0.12 0.32 -0.06 -0.44
(3.10) (11.82) (3.08) (1.48) (-1.71) (2.29) (-0.29) (-5.30)
Table VIII: Beta Expansion, Time-Series Analysis
This table examines time-series beta expansion associated with arbitrage trading. Panel A reports the baseline
regression. The dependent variable is the beta spread between the high-beta and low-beta deciles (ranked in
year π‘) in year π‘+1. πΆππ΅π΄π is the average pairwise partial weekly three-factor residual correlation within the
low-beta decile over the past 12 months. πΏππ£πππππ is a quintile dummy based on the average value-weighted
book leverage of the bottom and top beta deciles. We also include in the regression an interaction term
between πΆππ΅π΄π and πΏππ£πππππ. Panel B reports a battery of robustness checks. The dependent variable in
all rows is the beta spread between the high-beta and low-beta deciles in year π‘+1. Reported below is the
coefficient on the interaction of πΆππ΅π΄π and πΏππ£πππππ. Row 1 shows the baseline results which are also
reported in Table III. In Rows 2 and 3, we exclude the tech bubble crash and the recent financial crisis from
our sample. In Row 4, we also control for the UMD factor in computing πΆππ΅π΄π . In Row 5, we control for
both large- and small-cap HML in computing πΆππ΅π΄π . In Row 6, we control for the Fama-French five factor
model that adds profitability and investment to their three-factor model. In Row 7, we perform the entire
analysis on an industry-adjusted basis by sorting stocks into beta deciles within industries. In Row 8, we
instead measure the correlation between the high and low-beta portfolios, with a low correlation indicating
high arbitrage activity. In Rows 9 and 10, we examine upside and downside πΆππ΅π΄π , as distinguished by the
median low-beta portfolio return. In Rows 11-16, we replace πΆππ΅π΄π with residual πΆππ΅π΄π from a time-series
regression where we purge from πΆππ΅π΄π variation linked to, respectively, πΆππππ and πΆπππππ’π (Lou and
Polk, 2014), the average pair-wise correlation in the market, the lagged 36-month volatility of the BAB factor
(Frazzini and Pedersen, 2014), market volatility over the past 24 months, a trend, and lagged πΆππ΅π΄π (where
we hold the stocks in the low-beta decile constant but calculate πΆππ΅π΄π using returns from the previous year).
Standard errors are shown in brackets. *, **, *** denote significance at the 10%, 5%, and 1% level,
respectively.
Panel A: Baseline Regression
DepVar π΅ππ‘ππππππππ‘+1
[1] [2]
π΅ππ‘πππππππ 0.234*** 0.232***
[0.060] [0.058]
πΆππ΅π΄π 1.148** 0.163
[0.527] [0.636]
πΏππ£πππππ -0.035***
[0.015]
πΆππ΅π΄π β πΏππ£πππππ 0.408***
[0.117]
Adj-R2 0.083 0.103
No. Obs. 528 528
Panel B: Robustness Checks
DepVar = π΅ππ‘ππππππππ‘+1
Estimate Std. Dev.
Subsamples
Full Sample:1970-2016 0.408*** [0.117]
Excluding 2001 0.413*** [0.113]
Excluding 2007-2009 0.311** [0.140]
Alternative definitions of COBAR
Controlling for UMD 0.433*** [0.117]
Controlling for Large/Small-Cap HML 0.357*** [0.127]
Controlling for FF Five Factors 0.429*** [0.132]
Controlling for Industry Factors 0.473*** [0.128]
Correl btw High and Low Beta Stocks 0.142*** [0.040]
Upside πΆππ΅π΄π 0.513*** [0.172]
Downside πΆππ΅π΄π 0.281** [0.143]
Residual πΆππ΅π΄π
Controlling for πΆππππ 0.235** [0.115]
Controlling for πΆπππππ’π 0.240** [0.118]
Controlling for MKT CORR 0.382*** [0.122]
Controlling for Vol(BAB) 0.382*** [0.110]
Controlling for Mktvol24 0.261** [0.118]
Controlling for Trend 0.279** [0.119]
Controlling for Pre-formation πΆππ΅π΄π 0.433*** [0.120]
Table IX: Beta Expansion, Cross-Sectional Analysis
This table reports panel regressions of post-ranking stock beta on lagged πΆππ΅π΄π . At the end of each month,
all stocks are sorted into deciles based on their market beta calculated using daily returns in the past 12
months. To account for illiquidity and non-synchronous trading, we include on the right-hand side of the
regression equation five lags of the excess market return, in addition to the contemporaneous excess market
return. The pre-ranking beta is simply the sum of the six coefficients from the OLS regression. The dependent
variable is the post-ranking stock beta from year π‘ to π‘+1 (non-overlapping periods). The main independent
variable is lagged πΆππ΅π΄π , the average pairwise excess weekly return correlation in the low-beta decile over
the past 12 months. π·ππ π‘ππππ is the difference between a stockβs pre-ranking beta and the average pre-ranking
beta in year π‘. πΏππ£πππππ is the book leverage of the firm, measured in year π‘. We also include all double and
triple interaction terms of πΆππ΅π΄π , π·ππ π‘ππππ, and πΏππ£πππππ. Other (unreported) control variables include the
lagged firm size, book-to-market ratio, past one-year return, idiosyncratic volatility (over the prior year), and
past one-month return. Time-fixed effects are included in Columns 3 and 4. (Since πΆππ΅π΄π is a time-series
variable, it is subsumed by the time dummies.) Standard errors, shown in brackets, are double clustered at
both the firm and year-month levels. *, **, *** denote significance at the 10%, 5%, and 1% level, respectively.
DepVar πππ π‘ π ππππππ π΅ππ‘ππ‘+1 [1] [2] [3] [4]
πΆππ΅π΄π -0.873*** -0.823***
[0.154] [0.144]
π·ππ π‘ππππ 0.281*** 0.262*** 0.264*** 0.236*** [0.029] [0.032] [0.027] [0.030]
πΆππ΅π΄π β π·ππ π‘ππππ 0.768*** 0.571* 0.533** 0.531** [0.268] [0.298] [0.243] [0.268]
πΏππ£πππππ -0.002 -0.003* [0.002] [0.002]
πΆππ΅π΄π β πΏππ£πππππ 0.008 0.019 [0.016] [0.015]
πΏππ£πππππ β π·ππ π‘ππππ -0.004 0.005 [0.005] [0.005]
πΆππ΅π΄π β πΏππ£πππππ β π·ππ π‘ππππ 0.210*** 0.087** [0.050] [0.044]
Time Fixed Effects No No Yes Yes
Adj-R2 0.266 0.270 0.323 0.325
No. Obs. 1,194,641 1,194,641 1,194,641 1,194,641
Table X: Limits to Arbitrage
This table reports returns to the beta arbitrage strategy as a function of lagged πΆππ΅π΄π in various subsamples
ranked by proxies for limits to arbitrage (LTA). At the end of each month, all stocks are sorted into deciles
based on their market beta calculated using daily returns in the past 12 months. To account for illiquidity
and non-synchronous trading, we include on the right-hand side of the regression equation five lags of the
excess market return, in addition to the contemporaneous excess market return. The pre-ranking beta is
simply the sum of the six coefficients from the OLS regression. All months are then classified into five groups
based on πΆππ΅π΄π , the average pairwise partial return correlation in the low-beta decile over the past 12
months. Reported below is the difference in four-factor alpha to the beta arbitrage strategy between high
πΆππ΅π΄π periods and low πΆππ΅π΄π periods. Year zero is the beta portfolio ranking period. βLow LTAβ
corresponds to the subsample of stocks with low limits to arbitrage, and βhigh LTAβ corresponds to the
subsample with high limits to arbitrage. βLow-Highβ is the difference in monthly portfolio alpha between the
two subsamples. We measure limits to arbitrage using three common proxies. In columns 1-2, we rank stocks
into quartiles based on market capitalization (as of the beginning of the holding period); we label the top
quartile as βLow LTAβ and the bottom quartile as βHigh LTA.β In columns 3-4, we rank stocks into quartiles
based on idiosyncratic volatility with regard to the Carhart four-factor model (as of the beginning of the
holding period); we label the bottom quartile as βLow LTAβ and the top quartile as βHigh LTA.β In columns
5-6, we rank stocks into quartiles based on the illiquidity measure of Amihud (2002) (as of the beginning of
the holding period); we label the bottom quartile as βLow LTAβ and the top quartile as βHigh LTA.β T-
statistics, shown in parentheses, are computed based on standard errors corrected for serial-dependence up
to 24 lags. 5% statistical significance is indicated in bold.
Market Cap Idiosyncratic Volatility Illiquidity
Months 1-6 Year 3 Months 1-6 Year 3 Months 1-6 Year 3
Low LTA 0.82% -1.55% 0.93% -1.61% 0.82% -1.48%
(2.42) (-3.19) (2.74) (-3.84) (2.39) (-2.92)
High LTA -0.10% 0.10% 0.30% -0.14% -0.23% -0.11%
(-0.30) (0.18) (0.66) (-0.27) (-0.66) (-0.25)
Low-High 0.92% -1.65% 0.63% -1.47% 1.06% -1.38%
(2.41) (-2.82) (1.38) (-2.97) (2.73) (-3.03)
Table XI: Beta Arbitrage Timing Ability
This table reports regressions of monthly mutual funds and hedge funds returns on a conditional six-factor
model. At the end of each month, we track mutual funds and hedge fundsβ performance in the next six
months. The dependent variable in columns (1) and (2) is the average monthly excess return of long-short
equity hedge funds over that six-month window. Columns (3) and (4) are similar but use the monthly
excess returns on actively-managed mutual funds instead. We attribute those returns to πππ‘ππ, π ππ, βππ, π’ππ and π΅π΄π΅, the corresponding returns on the Fama-French three factors augmented by a momentum
factor and the four-factor alpha of the portfolio return buying stocks from the lowest beta decile and
shorting stocks from the highest beta decile. We allow the loading on BAB to be a function of lagged
πΆππ΅π΄π (quintile ranks from 1 to 5) and πππ§ππ πππ, the lagged cross-sectional ranking of the fundβs assets
under management. To compute πππ§ππ πππ, we first rank all funds (within the respective hedge fund or
mutual fund subset) into three groups as of the previous month and then assign the value 2 if the fund is
in the highest group, 1 if the fund is in the middle group, and 0 if the fund is in the lowest group. Standard
errors, shown in bracket, are clustered at both the month and fund level. *, **, *** denote significance at
the 10%, 5%, and 1% level, respectively.
Equity Hedge Funds Equity Mutual Funds
[1] [2] [3] [4]
πππ‘ππ 0.354*** 0.354*** 1.029*** 1.029*** [0.055] [0.055] [0.011] [0.011]
π ππ 0.241*** 0.241*** 0.216*** 0.216*** [0.039] [0.039] [0.021] [0.021]
βππ -0.049 -0.049 0.079*** 0.079*** [0.036] [0.036] [0.021] [0.021]
π’ππ 0.022 0.022 0.016* 0.016* [0.019] [0.019] [0.009] [0.009]
π΅π΄π΅ -0.102*** -0.127*** -0.026* -0.016 [0.022] [0.028] [0.014] [0.019]
πΆππ΅π΄π 0.001* 0.001** -0.000 -0.000 [0.000] [0.000] [0.000] [0.000]
π΅π΄π΅ β πΆππ΅π΄π 0.022*** 0.035*** 0.001 0.007 [0.008] [0.009] [0.005] [0.008]
πππ§ππ πππ -0.001** 0.000 [0.000] [0.000]
π΅π΄π΅ β πππ§ππ πππ 0.0248* -0.010 [0.014] [0.012]
πΆππ΅π΄π β πππ§ππ πππ -0.000 0.000 [0.000] [0.000]
π΅π΄π΅ β πΆππ΅π΄π β πππ§ππ πππ -0.013*** -0.006 [0.004] [0.004]
Adj-R2 0.018 0.018 0.690 0.690
No. Obs. 191,459 191,459 386,479 386,479
Figure 1: This figure shows the time series of the December observations of the πΆππ΅π΄π measure. At the end
of each month, all stocks are sorted into deciles based on their market beta calculated using daily returns in
the past 12 months. To account for illiquidity and non-synchronous trading, we include on the right-hand
side of the regression equation five lags of the excess market return, in addition to the contemporaneous
excess market return. The pre-ranking beta is simply the sum of the six coefficients from the OLS regression.
πΆππ΅π΄π is the average pairwise partial return correlation in the low-beta decile measured in the ranking period.
We begin our analysis in 1970, as that year was when the low-beta anomaly was first recognized by academics.
The time series average of πΆππ΅π΄π is 0.11.
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
0.16
0.18
0.2
πΆππ΅π΄π Time Series
Figure 2: This figure shows returns to the beta arbitrage strategy as a function of lagged πΆππ΅π΄π . At the end
of each month, all stocks are sorted into deciles based on their market beta calculated using daily returns in
the past 12 months. To account for illiquidity and non-synchronous trading, we include on the right-hand
side of the regression equation five lags of the excess market return, in addition to the contemporaneous
excess market return. The pre-ranking beta is simply the sum of the six coefficients from the OLS regression.
All months are then sorted into five groups based on πΆππ΅π΄π , the average pairwise weekly three-factor residual
correlation within the low-beta decile over the previous 12 months. The red curve shows the cumulative
Carhart four-factor alpha to the beta arbitrage strategy (i.e., a portfolio that is long the value-weight low-
beta decile and short the value-weighted high-beta decile) formed in high πΆππ΅π΄π periods, the dotted blue
curve shows the cumulative Carhart four-factor alpha to the beta arbitrage strategy formed in periods of low
πΆππ΅π΄π , and the grey curve shows the cumulative Carhart four-factor alpha to the beta arbitrage strategy
formed in periods of median πΆππ΅π΄π .
Cum
ula
tive
Four-
Fact
or
Alp
ha
Cum
ula
tive
Four-
Fact
or
Alp
ha
-5%
0%
5%
10%
15%
20%
25%
30%
-6 -4 -1 2 5 8 11 14 17 20 23 26 29 32 35 38 41 44 47 50 53 56 59
Low CoBar Median CoBar High CoBar
-10%
-5%
0%
5%
10%
15%
20%
25%
30%
0 2 5 8 11 14 17 20 23 26 29 32 35 38 41 44 47 50 53 56 59
Low CoBar Median CoBar High CoBar
Figure 3: This figure shows the security market line as a function of lagged πΆππ΅π΄π . At the end of each month,
all stocks are sorted into vigintiles based on their market beta calculated using daily returns in the past 12
months. To account for illiquidity and non-synchronous trading, we include on the right-hand side of the
regression equation five lags of the excess market return, in addition to the contemporaneous excess market
return. The pre-ranking beta is simply the sum of the six coefficients from the OLS regression. We then
estimate two security market lines based on these 20 portfolios formed in each period: one SML using portfolio
returns in months 1-6, and the other using portfolio returns in year 3 after portfolio formation; the betas used
in these SML regressions are the corresponding post-ranking betas. The Y-axis reports the average monthly
excess returns to these 20 portfolios, and the X-axis reports the post-ranking betas of these portfolios. Beta
portfolios formed in high πΆππ΅π΄π periods are depicted with a blue circle and fitted with a solid line, and those
formed in low πΆππ΅π΄π periods are depicted with a red triangle and fitted with a dotted line. The top panel
shows average excess returns and betas to the beta-arbitrage strategy in months 1-6 after portfolio formation,
while the bottom panel shows average excess returns and betas in year 3 after portfolio formation.
-4%
-3%
-2%
-1%
0%
1%
2%
0 0.5 1 1.5 2 2.5
Low CoBar High CoBar
-1%
-1%
0%
1%
1%
2%
2%
0 0.5 1 1.5 2 2.5
Low CoBar High CoBar
post-ranking beta
post-ranking beta
Aver
age
Exce
ss R
eturn
in M
onth
s 1-6
A
ver
age
Exce
ss R
eturn
in Y
ear
3
Figure 4: This figure shows how the relation between πΆππ΅π΄π and subsequent beta arbitrage returns varies
with firm leverage. At the end of each month, all stocks are sorted into deciles based on their market beta
calculated using daily returns in the past 12 months. To account for illiquidity and non-synchronous trading,
we include on the right-hand side of the regression equation five lags of the excess market return, in addition
to the contemporaneous excess market return. The pre-ranking beta is simply the sum of the six coefficients
from the OLS regression. All months are then sorted into five groups based on πΆππ΅π΄π , the average pairwise
weekly three-factor residual correlation within the low-beta decile over the previous 12 months. At the
beginning of the holding period, we sort stocks into four equal groups using book leverage. For each leverage
quartile, we compute the πΆππ΅π΄π return spread β i.e., the difference in Carhart four-factor alpha to the beta
arbitrage strategy (i.e., a portfolio that is long the value-weight low-beta decile and short the value-weighted
high-beta decile) between high and low πΆππ΅π΄π periods. The solid red curve shows the cumulative difference
in the πΆππ΅π΄π return spread between the highest and lowest leverage quartiles over the five years after
portfolio formation. This difference in the πΆππ΅π΄π return spread is 1.46%/month (t-statistic = 2.14) in year
one and is β0.52%/month (t-statistic = -2.41) in years four-five.
0%
5%
10%
15%
20%
25%
30%
0 1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 47 49 51 53 55 57 59
High-Leverage versus Low-Leverage Firms
Figure 5: This figure shows the conditional security market line as a function of lagged πΆππ΅π΄π (i.e., where
betas are allowed to vary with πΆππ΅π΄π ). At the end of each month, all stocks are sorted into vigintiles based
on their market beta calculated using daily returns in the past 12 months. To account for illiquidity and non-
synchronous trading, we include on the right-hand side of the regression equation five lags of the excess
market return, in addition to the contemporaneous excess market return. The pre-ranking beta is simply the
sum of the six coefficients from the OLS regression. We then estimate two conditional security market lines
based on these 20 portfolios: one SML using portfolio returns in months 1-6, and the other using portfolio
returns in year 3 after portfolio formation; the betas used in these SML regressions are the corresponding
post-ranking betas. The Y-axis reports the average monthly excess returns to these 20 portfolios, and the X-
axis reports the post-ranking beta of these portfolios. Beta portfolios formed in high πΆππ΅π΄π periods are
depicted with a blue circle and fitted with a solid line, and those formed in low πΆππ΅π΄π periods are depicted
with a red triangle and fitted with a dotted line. The top panel shows average excess returns and betas to
the beta arbitrage strategy in months 1-6 after portfolio formation, while the bottom panel shows average
excess returns and betas in year 3 after portfolio formation.
-4%
-3%
-2%
-1%
0%
1%
2%
0 0.5 1 1.5 2 2.5
Low CoBar High CoBar
-1%
-1%
0%
1%
1%
2%
2%
0 0.5 1 1.5 2 2.5
Low CoBar High CoBar
post-ranking beta
Aver
age
Exce
ss R
eturn
in M
onth
s 1-6
post-ranking beta Aver
age
Exce
ss R
eturn
in Y
ear
3
Figure 6: This figure shows how the information in πΆππ΅π΄π about time-variation in the expected holding and
post-holding return to beta-arbitrage strategies decays as staler estimates of beta are used to form the beta-
arbitrage strategy. At the end of each month, all stocks are sorted into deciles based on their market beta
calculated using daily returns in the past 12 months. To account for illiquidity and non-synchronous trading,
we include on the right-hand side of the regression equation five lags of the excess market return, in addition
to the contemporaneous excess market return. The pre-ranking beta is simply the sum of the six coefficients
from the OLS regression. We then compute the strategy return as the value-weight low-beta decile return
minus the value-weight high-beta decile return. We separately regress the abnormal return of the beta-
arbitrage strategy in months one-six and year three on πΆππ΅π΄π . In this process, we first use a fresh estimate
of beta, calculated using daily returns in the past 12 months. We then repeat the analysis using stale betas,
computed from daily returns in each of the prior 5 years (thus having different beta portfolios as of time zero
for each degree of beta staleness). We plot the corresponding regression coefficients (results for months 1-6
plotted with a blue square and results for year 3 plotted with a red circle) for each of the six beta-arbitrage
strategies, ranging from fresh beta to five years stale beta. The top (bottom) panel reports results based on
three-factor (four-factor) alpha.
Fam
a-F
rench
Alp
ha
Carh
art
Four-
Fact
or
Alp
ha
-0.25
-0.20
-0.15
-0.10
-0.05
0.00
0.05
0.10
0.15
0.20
0 1 2 3 4 5
Month 1-6 Year 3
Stale beta Fresh beta
-0.25
-0.20
-0.15
-0.10
-0.05
0.00
0.05
0.10
0.15
0.20
0 1 2 3 4 5
Month 1-6 Year 3
Fresh beta Stale beta